2026-02-14 11:47:26 +01:00
|
|
|
|
/**
|
2026-03-08 21:12:52 +01:00
|
|
|
|
* Environment variables parser.
|
2026-02-14 11:47:26 +01:00
|
|
|
|
* Parsed once at startup — all env access should go through this module.
|
|
|
|
|
|
*/
|
|
|
|
|
|
|
2026-03-08 21:12:52 +01:00
|
|
|
|
export interface Env {
|
|
|
|
|
|
IMMICH_API_URL: string;
|
|
|
|
|
|
IMMICH_API_KEY: string;
|
|
|
|
|
|
SITE_TITLE: string;
|
|
|
|
|
|
SITE_SUBTITLE: string;
|
|
|
|
|
|
CACHE_TTL: number;
|
Close the gaps left after v0.9.2's outage handling: route tests, stale cache, loading states (#397)
* docs: brainstorm resilience/adoption features and plan follow-up work
Brainstorm and deep-dive on where Immich Folio breaks under Immich
outages and what blocks self-hosted adoption, plus an implementation
plan for the remaining gaps after v0.9.2 (route-handler tests,
stale-while-error caching, loading skeletons).
* test: enable route-handler tests and cover admin guards and image content-type
vitest.config.ts only ran lib/__tests__/**, so route handlers were
structurally untestable. Two mechanisms live only at the route level and
cannot be extracted the way lib/imageSize.ts was: the admin auth guards
and the image proxy's Content-Type sanitisation (the stored-XSS fix from
c2fa8e7). Both were correct but unverified by any test.
Adds a table-driven test over all ten guarded /api/admin/* handlers,
including a filesystem walk that fails if a future route is added
without a row in the table. Adds a route-level test for
/api/image/[id] covering token validation, the six Content-Type
rewrite cases, ETag/304, and the 404 vs 503 (ImmichUnavailableError)
distinction.
Verified both catch real regressions by temporarily removing the admin
guards and the Content-Type rewrite and confirming the expected tests
fail, then restoring and confirming green again.
134 new tests (244 total, up from 210), tsc 0 errors, build succeeds.
* feat: serve stale cache entries while Immich is unavailable
v0.9.2 (#394) correctly split "gone" (404/410) from "unavailable"
(ImmichUnavailableError), and pages now render a 5xx instead of a false
404 during an outage — but that 5xx still replaced a gallery that had
loaded fine seconds earlier. lib/cache.ts deleted expired entries on
access, so nothing survived long enough to fall back to.
Cache entries now carry two deadlines: staleAt (the existing CACHE_TTL,
after which get() still reports a miss) and hardExpiresAt (the new
STALE_MAX_AGE, default 24h, after which the entry is dropped entirely).
Between the two, getStale() can still return the data.
getAlbums(), getAlbum() and getAssetInfo() catch ImmichUnavailableError
and fall back to getStale() before propagating. The two MISSING-sentinel
writes (definitive 404s) are deliberately left on the short cacheTtl
window — a corrected asset ID should take effect after the normal TTL,
not sit in the 24h stale window meant for outages.
STALE_MAX_AGE=0 disables the fallback. Verified the fallback only
engages on a real outage, not on a cache hit at the TTL boundary, and
that a definitive 404 still reports missing rather than falling back.
256 tests (up from 244), tsc 0 errors, build succeeds.
* feat: add loading skeletons for dynamic gallery routes
Every gallery route is force-dynamic and waits on a live Immich
round-trip — up to IMMICH_TIMEOUT_MS when the server is struggling —
with nothing rendered until it resolves. loading.tsx renders instantly
while the page component awaits its data.
Not browser-verified against live data: this worktree has no
.env.local and only .example content files, so the app boots into
setup mode (confirmed the dev server starts cleanly and the setup
screen renders, HTTP 200). tsc 0 errors, build succeeds, no new lint
errors.
* docs: document stale cache window, image token revocation, route-test convention
- README: STALE_MAX_AGE env var and a "Behaviour when Immich is
unreachable" section (stale-serves-until-STALE_MAX_AGE, 503 not 404,
outages never cached).
- docs/gallery-config.md: new "Image Token Revocation" subsection under
System & Security — un-publishing an album does not revoke already
issued image URLs; rotating AUTH_SECRET does.
- CLAUDE.md: updated cache description to cover the stale window and
MISSING sentinel, and a new code-convention note that route-level
logic (guards, header sanitisation) should get a route-level test
rather than being extracted into lib/ just to reach the test runner.
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
2026-08-07 11:07:34 +02:00
|
|
|
|
STALE_MAX_AGE: number;
|
fix: Immich outage handling, image tiers, admin save safety (#394)
* fix: migrate deprecated middleware convention to proxy
Next.js 16 deprecated the `middleware` file convention in favour of
`proxy`. The build warned on every run, and the convention is slated for
removal in a future major — at which point Next.js would silently stop
invoking this file. That is not a cosmetic failure: proxy.ts owns the
Content-Security-Policy and the per-request nonce, so an uninvoked file
means every document response ships with no CSP at all, with nothing in
the build output to catch it.
Renames middleware.ts to proxy.ts and the exported middleware() to
proxy(), per the official migration. The config.matcher export is
unchanged, so route coverage is identical.
Adds lib/__tests__/proxy.test.ts, which pins the export name and
verifies the behaviour a silent regression would take out: a CSP is set,
the nonce in the policy matches the one forwarded to pages via x-nonce,
nonces differ per request, x-pathname still reaches the root layout, and
the matcher continues to cover /admin.
Verified: tsc 0 errors, 105 unit tests passing (98 before), production
build succeeds with 0 deprecation warnings.
* fix: stop reporting Immich outages as missing albums
lib/immich.ts request() returned null for every failure mode: a 5xx, a
rejected API key, a network error, and a non-JSON body all looked exactly
like "Immich answered and said this does not exist". Callers had no way
to tell the two apart, and app/[...path]/page.tsx calls notFound() on a
null album in five places.
The result was that while Immich was down or returning 5xx, every album
URL served a hard 404 — for albums that exist. Visitors saw "not found",
and crawlers were free to de-index real content over a transient outage.
request() now throws ImmichUnavailableError for anything that is not a
definitive answer about the resource, and reserves null for 404 and 410.
Pages let that propagate, so an outage surfaces as a 500 rather than a
404. Two callers must not propagate and are handled explicitly:
- ping() catches and returns false — it answers "is Immich reachable",
and both callers (/api/health, /api/admin/status) render that as a
status field.
- /api/exif and /api/map return 503 with Retry-After instead of 404 and
an empty marker list respectively. An empty map is otherwise
indistinguishable from a gallery with no GPS data.
Adds 10 tests: each failure mode maps to the right outcome, getAlbum and
getAlbums propagate rather than reporting empty, ping stays non-throwing,
and a failed request leaves no stale entry in the in-flight coalescing
map — which would otherwise replay the failure after Immich recovered.
Verified: tsc 0 errors, 115 unit tests passing (105 before), production
build exit 0, eslint 0 errors on all touched files.
* fix: bound how long Immich requests may hang
lib/immich.ts set no timeout on any of its three fetch call sites, so the
only ceiling was undici's built-in default, measured in minutes. Every
page is force-dynamic, which means an Immich that accepts connections but
never answers held each visitor's render open for that whole window —
and since nothing caches the failure, load kept arriving.
Adds IMMICH_TIMEOUT_MS (default 15000, floored at 1000 so a stray 0
cannot mean "no timeout"), threaded through env.ts and AppConfig like
CACHE_TTL.
The two kinds of call need different treatment:
- request() carries a whole-exchange AbortSignal.timeout(). JSON
payloads are small, so one budget can cover headers and body.
- streamAsset()/streamVideo() use an AbortController whose timer is
cleared in a finally the moment response headers arrive. A large
original photo or a long video is legitimately slow to transfer; a
whole-request timeout would truncate playback and break seeking, since
every range request would restart the clock.
A timed-out request surfaces as ImmichUnavailableError naming the budget,
so a misconfigured value is diagnosable from the log line alone.
Adds 5 tests, including the two that pin the distinction: a stream whose
headers have arrived is not aborted an hour later, while a stream that
never sends headers still is.
Verified: tsc 0 errors, 120 unit tests passing (115 before), production
build exit 0, no new eslint errors (the one error in env.ts is
pre-existing — confirmed against a clean tree).
* fix: stop caching Immich outages as permanently missing media
f3 split "gone" from "unavailable" in request(), but streamAsset() and
streamVideo() have their own fetch calls and were left returning null for
every failure. Both proxy routes turn null into a 404.
That 404 carried no Cache-Control, and a 404 is heuristically cacheable
per RFC 9111 — on URLs the success path serves as `immutable`. So a
transient outage, or a timeout introduced in 370e668, could pin a broken
image or an unplayable video in the visitor's browser cache well after
Immich recovered.
Both stream methods now throw ImmichUnavailableError for transport
failures, timeouts, and any error status other than 404/410, and reserve
null for genuinely absent media. A 200 with no body counts as
unavailable, not as missing. The routes translate that into 503 with
Retry-After, and both error responses — the 503 and the real 404 — now
send Cache-Control: no-store.
streamVideo keeps accepting 206, so range requests and seeking are
unaffected.
Adds 7 tests covering each failure mode for both methods plus the 206
path, and updates the F4 timeout test, which asserted the old null return.
Verified: tsc 0 errors, 127 unit tests passing (120 before), production
build exit 0, eslint 0 errors on touched files.
* perf: cache "does not exist" answers from Immich
getAlbum() and getAssetInfo() cached successes but discarded a definitive
"not found", so a missing resource was re-queried upstream on every call.
Measured before: 5 sequential lookups of a missing asset produced 5
upstream fetches. After: 1.
This is not hypothetical. app/page.tsx resolves every gallery.yaml hero
ID through getAssetInfo() on each render, and the homepage is
force-dynamic — so a hero photo deleted from Immich but still listed in
gallery.yaml cost one 404 round-trip per page view, indefinitely.
Absence is stored as a frozen MISSING sentinel because the cache cannot
distinguish a miss from a stored null; both read back as null.
Only definitive 404/410 answers are cached. An outage throws
ImmichUnavailableError and is deliberately never stored — caching one
would keep the gallery broken after Immich recovered. A test pins that:
a 503 followed by a healthy response must reach Immich the second time.
Stored under the normal cacheTtl rather than a separate shorter negative
TTL, because the reason for a shorter one no longer applies: outages are
not cached at all now, and both invalidation paths (the Immich webhook
and the admin save/reload) clear the entry immediately, so a corrected
asset ID takes effect at once rather than waiting out a TTL.
Also raises the test config's cacheTtl from 0, which expired entries in
the same millisecond they were written and would have made any caching
assertion flaky. Test isolation comes from cache.clear() in beforeEach.
Verified: tsc 0 errors, 131 unit tests passing (127 before), production
build exit 0, eslint 0 errors on touched files.
* fix: add themed error and not-found pages
The app had no app/error.tsx, app/not-found.tsx or global-error.tsx, so
every failure fell through to Next.js's unstyled default page.
That gap became visible with 253ade0, which made an Immich outage produce
a 500 instead of a false 404 — correct, but what a visitor saw was a bare
framework error screen with none of the site's theming.
The two pages also carry the distinction the error split was for, and
they carry it structurally rather than by inspecting the error: notFound()
renders not-found.tsx, a throw renders error.tsx. That matters because
Next.js strips error messages in production and replaces them with a
digest — this component genuinely cannot tell an Immich outage from any
other server error, so it does not claim to. It frames the failure as
temporary and offers reset(); not-found.tsx deliberately does not, since
a real 404 is not transient.
The digest is shown so a user report can be traced to a log line. The raw
error message is never rendered — it can carry an upstream URL, a header
or a config value.
Both pages reuse .empty-state, which existed in globals.css but was
unreferenced anywhere in the app; extended with title, actions and button
elements matching the PasswordGate button, the only other action button
in the codebase.
Tested by server-rendering both components (react-dom/server, no new
dependencies) — 8 tests covering the styling hook, the retry and home
affordances, digest handling, and that a message containing an API key or
host address does not reach the markup. Note vitest.config.ts only
includes lib/__tests__/**/*.test.ts, so the test uses createElement
rather than JSX; a .tsx file there would never run.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the app boots into setup mode and neither
Playwright nor a dev server can exercise these routes here.
Verified: tsc 0 errors, 139 unit tests passing (131 before), production
build exit 0 with /_not-found registered, eslint 0 errors.
* perf: drop the unused originalUrl from every photo item
PhotoItem declared originalUrl and toPhotoItems computed it for every
asset, but a repo-wide grep found exactly two references: the field
declaration and the assignment. Nothing ever read it — the grid uses
thumbUrl, the lightbox uses previewUrl.
The cost was not the encryption. PhotoItem is a server-to-client prop, so
the dead field was serialized into the RSC payload and shipped to the
browser for every photo.
Measured: 137 bytes per asset once serialized, so 6.7 KB for a 50-photo
album and 66.9 KB for a 500-photo one — transferred and parsed on every
page view, never read.
Removed rather than wired up to a "view original" affordance. Exposing
full-size originals is a bandwidth and privacy decision for the project
to make deliberately, not something to infer from a dangling field.
No unit test: this is a deletion whose correctness is exactly "nothing
referenced it", which tsc and a repo-wide grep already establish. A test
asserting the absence of a field would restate the type checker.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint unchanged at 4 pre-existing errors on the touched files (unused
imports, confirmed identical against a stashed clean tree).
* chore: remove four orphaned imports from the catch-all route
Link, Image, PhotoGrid and BackLink were imported by app/[...path]/page.tsx
and never used — 4 of the 6 no-unused-vars errors in the repo baseline.
Verified they are leftovers, not a dropped render path: PhotoGrid and
BackLink are both rendered by AlbumDetailView, which imports them itself,
and BackLink is also used by the impressum and map pages. Rendering was
extracted out of page.tsx into AlbumDetailView/SubpageGridView at some
point and the imports stayed behind. `type PhotoItem` is still needed by
toPhotoItems, so that import narrows to a type-only import rather than
disappearing.
No bundle win, measured: .next/static is 1084 KB across 20 JS chunks both
before and after, byte-identical. Turbopack already tree-shook them. The
backlog note that predicted unused client-component imports would hold
them in the client graph was wrong, and has been corrected there.
The value is readability and lint: repo errors drop from 54 to 50, and
page.tsx no longer reads as though it renders the grid itself.
chore: rather than fix:/perf: — nothing behaves differently and the
measurement showed no performance change, so either of those labels would
misdescribe it. CLAUDE.md lists chore: as a valid prefix.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint 0 errors on the touched file.
* fix: make ?w= able to narrow the image tier instead of being ignored
The image proxy resolved its Immich rendition with `?size=` winning
whenever present, falling back to `?w=` otherwise. lib/urls.ts always
writes `?size=`, so for every URL this app generates the width branch was
unreachable and widthToSize() was dead code — while CLAUDE.md documented
the width mapping as the actual behaviour.
`?size=` is now a ceiling rather than an override: when both are present
the smaller tier wins, so a width can narrow the request but never widen
it. That asymmetry is the point. Letting width win outright — the obvious
reading of the docs — would have been a real regression: next/image emits
widths up to 3840 and widthToSize(1920) is 'original', so every large
display would have started downloading full-size originals in place of
previews. A test pins that specific case.
No measured change today. The only URLs carrying both parameters are
?size=preview with next/image widths, whose smallest vw-based deviceSize
is 640 — already preview either way. This is a correctness fix that makes
the documented contract true and testable, not a performance win.
Moved the logic to lib/imageSize.ts so it can be tested at all:
vitest.config.ts only collects lib/__tests__, so nothing inside a route
file is reachable from the suite. 9 tests cover the tier boundaries, both
single-parameter paths, malformed input, and the ceiling behaviour.
CLAUDE.md updated to describe the ceiling rule rather than a plain
width mapping.
Verified: tsc 0 errors, 148 unit tests passing (139 before), production
build exit 0, eslint 0 errors on touched files.
* fix: validate the per-album heroImage UUID from gallery.yaml
Every asset ID in gallery.yaml is routed through validateUuid before it
can reach encodeAssetId, which encrypts whatever it is handed. One was
not: for a per-album entry the album ID (the map key) was validated while
the heroImage value was stored raw.
A typo there produced a perfectly well-formed image URL that 404s forever,
with nothing in the log naming the cause — the failure surfaced as a
broken hero image on an album page, far from the line that caused it.
This is narrower than the finding that prompted it, which proposed
guarding encodeAssetId itself. Validation belongs at config load, where it
runs once and can name the file and the album; a guard in encodeAssetId
would run on every render and duplicate validateUuid at the wrong layer.
The heroImage path was the only unvalidated route into it.
validateUuid warns and substitutes the nil UUID rather than throwing, so a
bad ID still degrades to a missing image instead of taking the page down —
and since f511e62 that 404 is negative-cached, so it costs one upstream
request rather than one per render.
Adds 4 tests. The two that pin the wiring were mutation-checked: both fail
against the unfixed code and pass with it.
Verified: tsc 0 errors, 152 unit tests passing (148 before), production
build exit 0, eslint 0 errors on touched files.
* fix: keep /admin reachable when gallery.yaml cannot be derived
getConfig() throws on three shapes the admin page builder is able to
write: a gallery with neither albums nor subpages, a subpage with no name,
and a subpage with neither albums nor sections.
RootLayout called it unguarded, and a throw in a *layout* is not caught by
app/error.tsx — that needs a global-error boundary, which this app does
not have. /admin renders inside that same layout, so one bad save took
down the site together with the only tool that could undo it. Recovery
meant editing content/gallery.yaml on the host, or exec'ing into the
container. This is issue #326's failure mode reached through an invalid
config instead of a missing one.
Adds getConfigOrNull(), which returns null rather than throwing and logs
the underlying error plus the path to content/.backups/. The layout treats
null exactly like needsSetup: setup screen for visitors, children rendered
for admin paths so the panel still loads. generateMetadata() is guarded
too — it also calls getConfig(), and it also runs before any error
boundary.
This is the half of the finding that makes every broken config
recoverable, including ones this app never wrote: a hand-edited file, a
restored backup, a config from an older version. The other half —
validating in the admin PUT so the bad state is never written — is filed
separately; it prevents new breakage but cannot repair existing breakage,
so it was not the one to do first.
Adds 10 tests. Three pin the premise by asserting getConfig() really does
throw on each shape, so the fallback cannot quietly become dead weight.
One asserts the needsSetup path still returns a config rather than null,
since missing and invalid are different states that must not collapse.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the layout/error-boundary interaction is
reasoned from Next.js semantics and covered at the getConfig level only.
Verified: tsc 0 errors, 162 unit tests passing (152 before), production
build exit 0, eslint 0 errors on touched files.
* fix: reject an admin save the site would not be able to load
The admin PUT validated only the shape of the hero field, then wrote. None
of getConfig()'s three throw conditions was checked, so the page builder
could persist a gallery.yaml that cannot be loaded — and the user got
"Saved successfully" for it. Since 95877af that no longer locks anyone
out, but the public site still drops to the setup screen until someone
notices.
Extracts the gallery half of getConfig() into deriveGallery(), which the
route now runs against the proposed gallery before writing. Two
alternatives were rejected on purpose:
- Re-checking the three conditions in the route would put a second copy of
the rules next to the first, and the two would drift.
- Write-then-verify-then-roll-back leaves a window in which other requests
read the broken config.
The extraction is a pure move: no logic changed inside it, and all 162
pre-existing tests pass untouched, which is what makes that claim
checkable rather than asserted.
The 400 carries the derivation's own message, so the admin UI names the
offending subpage instead of saying "invalid gallery" — a builder with a
dozen subpages needs to know which one. A test pins that the message
identifies it.
Adds 8 tests: each rejected shape, the named-subpage message, and the
valid structures (standalone albums, subpage membership, sections,
per-album overrides and hero images) so the extraction cannot silently
lose behaviour.
Verified: tsc 0 errors, 170 unit tests passing (162 before), production
build exit 0, eslint 0 errors on touched files.
* fix: hash subpage passwords off the main thread
verifyScrypt used crypto.scryptSync, measured at 22.8ms per call — 22.8ms
during which the Node process serves nothing else, because the sync
variant runs on the main thread rather than the libuv threadpool.
The /api/auth rate limit does not contain this. It allows 10 attempts per
minute, but the bucket key is the client IP, and with the default
TRUSTED_PROXY_HOPS=0 that IP is read from x-real-ip / x-forwarded-for —
both attacker-controlled. lib/rate-limit.ts already documents that those
headers are spoofable at hops=0; what followed from it was not written
down anywhere: one attacker rotating a header gets unlimited buckets, and
each request buys 22.8ms of dead process. Roughly 44 requests per second
saturates it.
Switches verifyScrypt and generateScryptHash to the async crypto.scrypt,
which runs on the threadpool. This does not stop the spoofing — that is
inherent to hops=0 and already documented — it removes the amplification
that made spoofing worth doing.
authenticate() becomes async. It had exactly one caller
(app/api/auth/route.ts). isAuthenticated() is deliberately left
synchronous: it only computes an HMAC, and it is called from server
components.
Adds 4 tests, two of which pin the fix and were mutation-checked against
the old implementation — both fail with scryptSync and pass with scrypt.
One asserts scryptSync is never called; the other asserts a 2ms interval
still fires during a hash, which is the property that actually matters. A
slower machine lengthens the hash and yields more ticks, so that test
fails on a blocked loop rather than on a busy one.
Also adds a scrypt-hashed fixture subpage, so the suite exercises
verifyScrypt rather than only the deprecated plaintext fallback.
Verified: tsc 0 errors, 174 unit tests passing (170 before), production
build exit 0, eslint 0 errors on touched files.
* fix: warn when TRUSTED_PROXY_HOPS cannot identify a client
getClientIp falls back to a shared 'unknown' bucket whenever the
X-Forwarded-For chain is shorter than the configured hop count, or the
header is absent with hops > 1.
That fallback is correct — the alternative is trusting a spoofable entry —
but it was silent, and silence is the problem. Every unidentified request
shares one bucket per endpoint, and this is an image proxy: a single
50-photo grid issues ~50 /api/image requests, so the default
RATE_LIMIT_RPM=1500 is collectively spent by roughly 30 page loads a
minute. Someone who sets TRUSTED_PROXY_HOPS=2 for Cloudflare-plus-nginx
and then reaches nginx directly gets a site that 429s everyone, with
nothing in the log pointing at the cause.
Warns once per process, naming the configured hop count and what was
actually seen — the chain length, or which header was missing. Once, not
per request: enough to diagnose without flooding the log.
The fallback itself is unchanged. Trusting a shorter chain would hand
every client its own bucket key and defeat the limiter entirely.
No warning at hops=0, where the shared bucket is the documented
consequence of not declaring a proxy rather than a misconfiguration.
Adds 6 tests covering each fall-through path, the once-per-process
behaviour, and both silence cases. The once-per-process flag needs a test
seam to re-arm, hence the __resetProxyWarningForTests export.
Verified: tsc 0 errors, 180 unit tests passing (174 before), production
build exit 0, eslint 0 errors on touched files.
* fix: send Retry-After on every rate-limited response
/api/image, /api/video and /api/exif each computed the retry delay,
interpolated it into a console.warn, and then returned a 429 without it —
the number was calculated and thrown away. /api/og did not compute it at
all. Only /api/auth and /api/map sent the header.
This matters more for an image proxy than for a typical API: one gallery
page issues around fifty /api/image requests, so a burst that trips the
limit produces fifty rejections carrying no back-off signal. The browser
and next/image then retry on their own schedule — a retry storm against a
server that just reported it was overloaded.
Adds Retry-After and Cache-Control: no-store to all four. no-store because
the success path serves these same URLs as `immutable`.
The delay now comes from a shared retryAfterSeconds() helper rather than
an inline Math.ceil in each route, which also fixes an edge every copy
shared: a window on the point of expiring yields 0, and Retry-After: 0
tells the client to retry immediately, which is the opposite of what a 429
means. A resetAt already in the past yields a negative number, which is
not valid HTTP. The helper floors at 1.
The three routes that already sent the header (/api/auth, /api/map,
/api/admin/auth) and /api/health now use the helper too — leaving four
copies of the buggy arithmetic behind while centralising the rest would
have been worse than not centralising at all.
Adds 2 tests, one of them specifically for the zero-and-negative edge.
Verified: tsc 0 errors, 182 unit tests passing (180 before), production
build exit 0, eslint 0 errors on all ten touched files.
* fix: add an escape hatch for stale browser image caches
Image and video responses are served
`Cache-Control: public, max-age=31536000, immutable`, so a browser does
not revalidate them for a year — the ETag is not even consulted. When
Immich regenerates thumbnails after a job re-run, or a photo is rotated,
the asset ID does not change, so neither does the deterministic token or
the URL. Visitors keep the old rendition, and the webhook that clears the
server-side caches cannot reach a browser. The only existing remedy was
rotating AUTH_SECRET, which invalidates every token on the site.
Corrects the framing this started from. The finding described it as an
ETag problem, but `immutable` means the ETag is never reached in the case
that matters. The URL is the only lever a browser cache responds to, so
the buster belongs there. It is also mixed into the ETag, since a matching
ETag across two different URLs would be wrong regardless.
Adds IMAGE_CACHE_VERSION, empty by default. Unset, the generated URLs are
byte-identical to before — a test asserts exactly that, because a
cache-busting feature that changes URLs when nobody asked for it would
silently discard every visitor's cache on deploy.
The value is URL-encoded at parse time; an unescaped one could otherwise
inject query parameters into every image URL on the site.
Deliberately not done: shortening max-age or dropping `immutable`. Both
trade a large, real bandwidth win for a rare correctness case, and the
cost is not measurable from this worktree. The aggressive caching stays
the default; this only adds a way out.
Adds 6 tests: default-unchanged, whitespace treated as unset, both URL
shapes, escaping, and that a bump changes the query while leaving the
token intact.
Verified: tsc 0 errors, 188 unit tests passing (182 before), production
build exit 0, eslint unchanged (the one error in env.ts is pre-existing,
confirmed against a stashed tree).
* security: validate the backup filename before restoring from it
restoreBackup() joined its argument straight into a path and copied the
result over a content file. path.join resolves `..`, so
restoreBackup('../../../../etc/passwd') would have written any readable
file over content/settings.yaml — whose contents the admin GET endpoints
hand straight back. The destination was chosen by
backupFilename.includes('gallery.yaml'), which is substring matching on
the same attacker-supplied string, so the caller picked that too.
Not exploitable today: the function has zero callers. That is precisely
why it was worth fixing rather than leaving — it is an exported function
carrying a traversal bug, waiting for whoever wires up the restore button
the admin panel does not yet have.
Hardened rather than deleted. The status endpoint already reports backup
counts, so restore is plainly intended; removing a half-built feature is
the maintainer's call, not a side effect of a security fix.
The name must now match the pattern this service itself produces, and the
destination is derived from the matched group rather than a substring
search. The redundant basename check is kept on purpose: it preserves the
guarantee if that pattern is ever loosened.
Adds 12 tests. Nine of them fail against the unguarded version —
mutation-checked by removing the guard and confirming, rather than
assuming they would.
Verified: tsc 0 errors, 200 unit tests passing (188 before), production
build exit 0, eslint 0 errors on touched files.
* fix: give each admin save its own temp file
writeYamlFile wrote to a constant `${filePath}.tmp` before renaming. The
rename is what makes the write atomic; the temp filename was not. Two
saves of the same file in flight at once shared that path, so the second
writeFile could interleave with the first's rename — publishing one save's
bytes under the other's, or a truncated mix of both. A double-clicked Save
in the page builder is enough to reach it.
The temp name now carries the pid and a per-process counter, so concurrent
writers cannot collide either within a process or across workers.
Also wraps the write and rename so a failure unlinks the temp file instead
of leaving it in content/. The cleanup swallows its own errors: failing to
tidy up must not replace the real error with a misleading one.
Adds 7 tests, and both halves were mutation-checked independently —
reverting to the fixed temp path fails two of them, removing the cleanup
fails two others. Two of the tests exist to pin properties that are easy
to regress silently: that the rename targets the temp file just written,
and that the happy path issues no unlink.
Verified: tsc 0 errors, 207 unit tests passing (200 before), production
build exit 0, eslint 0 errors on touched files.
* fix: await the stream-timeout rejection before advancing timers
`npm run test:unit` exited 1 even though all 207 tests passed. The
timeout test started `streamAsset`, advanced the fake timers past the
15s budget, and only then attached the rejection expectation. The abort
fires *during* `advanceTimersByTimeAsync`, so for one microtask turn the
promise was rejected with no handler attached — Node reports that as an
unhandled rejection, and vitest fails the whole run on it regardless of
test outcomes. CI would have blocked on a green suite.
Attaching the assertion before advancing the timers fixes the ordering
without weakening it: the same rejection and the same `captured.aborted`
check still run, just awaited after the advance rather than created
after it.
Verified: vitest exit 0, 207/207 passing, tsc 0 errors, prettier clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-05 12:04:01 +02:00
|
|
|
|
IMAGE_CACHE_VERSION: string;
|
|
|
|
|
|
IMMICH_TIMEOUT_MS: number;
|
2026-03-08 21:12:52 +01:00
|
|
|
|
RATE_LIMIT_RPM: number;
|
|
|
|
|
|
AUTH_SECRET?: string;
|
2026-07-20 18:57:57 +02:00
|
|
|
|
TRUSTED_PROXY_HOPS: number;
|
feat: add visual admin panel with page builder and settings editor
- Add /admin route with password-protected visual page builder
- Page Builder: manage hero images, standalone albums, subpages, and sections
- Album Picker: browse all shared Immich albums with search and filtering
- Settings Editor: theme, grid, footer, legal, SEO configuration
- YAML service with atomic writes and automatic backups (content/.backups/)
- HMAC-signed admin sessions (24h expiry, HttpOnly cookies)
- API endpoints: /api/admin/{auth,gallery,settings,albums,reload}
- Enable with ADMIN_PASSWORD env var; disabled by default
- invalidateConfigCache() export for hot-reload after saves
2026-05-29 17:04:37 +02:00
|
|
|
|
WEBHOOK_SECRET?: string;
|
|
|
|
|
|
ADMIN_PASSWORD?: string;
|
release: v0.11.0 — journal fixes, admin URLs, design system cleanup (#432)
* fix(screenshots): sharp-Typen fuer 0.34 und 0.35 gleichzeitig (#410)
sharp 0.35 ist von `export =` auf ESM umgestellt: der Modul-Namespace
ist nicht mehr aufrufbar, die Factory sitzt unter .default. Der
Type-Check brach damit in scripts/screenshots.ts weg, sobald der
Next-Bump (#402) sharp von 0.34.5 auf 0.35.3 mitzieht -- sharp ist hier
keine deklarierte Dependency, sondern kommt ueber Next mit.
Die Form wird jetzt aus dem installierten Paket abgeleitet statt
festgeschrieben, zur Laufzeit wie im Typ. Geprueft mit beiden
Versionen: tsc laeuft mit 0.34.5 und mit 0.35.3 fehlerfrei durch.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore(deps): bump the nextjs group across 1 directory with 2 updates (#402)
Bumps the nextjs group with 2 updates in the / directory: [next](https://github.com/vercel/next.js) and [eslint-config-next](https://github.com/vercel/next.js/tree/HEAD/packages/eslint-config-next).
Updates `next` from 16.2.12 to 16.3.0
- [Release notes](https://github.com/vercel/next.js/releases)
- [Commits](https://github.com/vercel/next.js/compare/v16.2.12...v16.3.0)
Updates `eslint-config-next` from 16.2.12 to 16.3.0
- [Release notes](https://github.com/vercel/next.js/releases)
- [Commits](https://github.com/vercel/next.js/commits/v16.3.0/packages/eslint-config-next)
---
updated-dependencies:
- dependency-name: eslint-config-next
dependency-version: 16.3.0
dependency-type: direct:development
update-type: version-update:semver-minor
dependency-group: nextjs
- dependency-name: next
dependency-version: 16.3.0
dependency-type: direct:production
update-type: version-update:semver-minor
dependency-group: nextjs
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* chore(deps-dev): bump prettier from 3.8.3 to 3.9.6 (#405)
Bumps [prettier](https://github.com/prettier/prettier) from 3.8.3 to 3.9.6.
- [Release notes](https://github.com/prettier/prettier/releases)
- [Changelog](https://github.com/prettier/prettier/blob/main/CHANGELOG.md)
- [Commits](https://github.com/prettier/prettier/compare/3.8.3...3.9.6)
---
updated-dependencies:
- dependency-name: prettier
dependency-version: 3.9.6
dependency-type: direct:development
update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* docs(changelog): Unreleased-Sektion fuer die Bumps nach v0.10.0 (#413)
Die Dependency-PRs sind nach dem Release auf dev gelandet, teils auch
auf main. Ohne Eintrag waeren sie beim naechsten Release unsichtbar --
inklusive Next 16.3.0, das CVE-2025-13465 mitbringt.
Ausserdem im 0.10.0-Abschnitt nachgetragen, dass der Trivy-Scan bis
#409 nie gelaufen ist. Der Fix steckt im Tag v0.10.0, fehlte aber in der
Aufzaehlung -- und "Images werden bei jedem Publish gescannt" stand im
Changelog, waehrend der Job im Action-Setup abbrach.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(gallery): per-album sort order, independent of Immich (#408) (#414)
Album order was dictated entirely by Immich: getAlbum() sorted by the
album's own asc/desc flag, so a Folio album always mirrored the archive.
A curated series has a narrative order that has nothing to do with
capture time, and changing the Immich album's sort to fit the website
changes it for the archive too.
Adds an optional per-album `sort` in gallery.yaml, editable in /admin,
defaulting to today's behaviour so every existing config is unchanged:
immich (default) | newest | oldest | filename | manual
`manual` is a pinned *prefix*: `assetOrder` lists only the photos placed
by hand, everything else follows in the album's Immich order. A full
permutation of a 2,000-photo album would be ~80 KB of UUIDs in a file
that is rewritten and copied to content/.backups/ on every admin save.
It also makes drift fall out for free — photos removed in Immich never
match, new ones land in the tail rather than being hidden.
Sorting is applied after the cache read, not before the cache write, so
the LRU keeps one canonical Immich-ordered copy per album. That avoids
putting the mode in the cache key, lets a config change take effect
without depending on invalidation, and keeps the stale-serving path
correct. cache.get() returns the stored object by reference, so
withSort() returns a shallow clone with a new array — sorting in place
would reorder the shared entry, and under request coalescing every
concurrent caller would be sorting the same array.
Also fixes a prerequisite bug: /api/admin/albums/[albumId]/assets read
album.assets from GET /albums/:id, which Immich 3.x no longer populates
(lib/immich.ts:507). The hero picker's "This Album" tab returned nothing
and threw outright when the key was absent. It now goes through the
client's paginated metadata search, with ?types=all to opt videos in for
the reorder editor.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): drop decorative emoji from the sort select labels (#415)
An <option> cannot contain markup, so the SVG set in Icons is not
reachable there — but a decorative emoji is not a substitute for it.
Plain text instead; the album card badge already uses a real icon.
The remaining emoji in the admin panel (asset picker tabs, section
headings, layout picker options) predate this branch and are left for a
separate cleanup.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): replace emoji icons with the shared SVG icon set (#416)
The admin panel mixed two icon conventions: inline SVGs from
`Icons.tsx` in most components, and decorative emoji in the asset
picker and a few PageBuilder labels. Emoji render inconsistently
across platforms, ignore the accent colour, and cannot inherit the
`.svg-icon` sizing rules.
- Icons.tsx: add `IconStar` (filled — it marks a state, not an
action) and `IconImage`.
- AssetPicker: the three tab labels and the favourite tile badge now
use IconFolder / IconStar / IconImage.
- PageBuilder: the "Standalone Albums" and "Subpages" headings use
the icon set instead of an emoji.
- PageBuilder: the layout picker drops its emoji for plain text. An
<option> cannot contain markup, so an SVG is impossible there and a
decorative glyph is not a substitute.
- admin.css: align icon and label in the picker tabs, and swap the
favourite badge's text-shadow for drop-shadow, which is what
applies to an inline SVG.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): fold PageBuilder's local icons into the shared Icons module (#418)
* refactor(admin): fold PageBuilder's local icons into the shared module
PageBuilder.tsx carried its own `Icons` object duplicating most of the
glyphs already exported by Icons.tsx, with hardcoded sizes and slightly
divergent paths. Migrate all 32 call sites to the shared module and
delete the local object.
Five glyphs had no counterpart and are added to Icons.tsx as 1:1 ports
of the local paths: IconX, IconPlus, IconHome, IconCopy, IconImage.
The local ExternalLink and Check are dropped rather than ported — both
had zero call sites.
Two glyphs change shape slightly, which is the point of consolidating:
the shared IconCamera includes the viewfinder bump the local copy
omitted, and drag handles now use IconGripVertical.
IconGripVertical uses the module's square 24x24 viewBox, so `.svg-drag`
goes from 12x18 to 18x18 — the old non-square box would letterbox the
glyph down to 12x12 and shrink the handle.
* fix(admin): resolve icon-module integration after #416
Merging dev brought in #416, which added a second IconImage to Icons.tsx
and two <Icons.Camera>/<Icons.Folder> section headers that reference the
local Icons object this PR removes. Git merged both cleanly but the result
did not type-check.
- drop the duplicate IconImage (both copies were byte-identical)
- point the two section headers at the imported IconCamera/IconFolder
- refresh the SORT_LABELS comment: the icon set is imported now, not
declared below
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* fix(config): allow an empty gallery, guard against a double /api suffix (#421)
An empty gallery is now a valid, rendered state instead of a throw: deriveGallery() no longer requires at least one album or subpage, so a deployment with no albums yet renders a bare homepage the owner can fill in via /admin.
getConfig() also stops appending /api to an Immich URL that already ends in it, which previously produced .../api/api and broke every upstream request.
Includes a restore of the per-album sort test cases, which the branch had lost by being cut before #408/#414.
* feat(admin): favicon upload, served from the content volume (#422)
Adds a bundled default favicon, an upload widget in admin Settings > General, and a route that serves the icon from the writable content/ volume with a fallback to the bundled default. Raster uploads (PNG/ICO/JPEG) are wrapped in an SVG container; the response carries its own CSP so an uploaded SVG cannot execute script on the origin.
The GET route revalidates with an ETag instead of caching immutably, so a newly uploaded icon shows up immediately. Uploads are written via temp file + rename, and a Reset button restores the bundled default.
* ci: flag unformatted files in a PR, without blocking it (#424)
Prettier is configured as an error rule in eslint.config.mjs but has never run in CI, so part of the tree predates it. A final CI step now diffs against the PR base, runs prettier --list-different over the changed files, and emits a warning annotation per offender plus a job summary.
Advisory by design: it never fails the job. Blocking would turn a PR red for merely touching a file that was already unformatted, and normalising the tree in one commit would wreck git blame and conflict with the PRs in flight.
Requires fetch-depth: 0, since diffing against the base needs its history.
* feat(install): first-run wizard gated by a one-time setup token (#419)
Adds /install, a four-step first-run wizard: connect to Immich, optionally pick albums, name the site, finish. It verifies the credentials against the Immich server before writing anything, then writes gallery.yaml, settings.yaml and a content/install.json carrying the Immich credentials and a generated site secret.
install.json exists because a running process cannot pick up new env vars — the wizard has to write configuration that takes effect without a restart. getInstallCredentials() merges it under the real environment, so env vars always win and rotation stays a matter of setting one.
Both the page and the API routes are gated by a one-time setup token printed to the server log, so a deployment that is reachable before gallery.yaml exists cannot be configured by whoever finds it first. The token is stored in content/.setup-token (0600) rather than a module variable: Next.js bundles the page and each route handler separately, so a module-level token left the page and the API disagreeing inside one process and the wizard stuck on step 1. It is removed once the install completes.
The admin password is stored as an scrypt hash using the scheme lib/auth.ts already applies to album passwords; a plaintext ADMIN_PASSWORD env var keeps working.
* docs: document the setup wizard, fix the read-only content mount (#425)
Quick Start now leads with the /install wizard, with the hand-written YAML route kept alongside it, and a First-Run Setup section covers the setup token, the files the wizard writes, and env-var precedence.
Two corrections: both Docker examples mounted content/ as :ro, which silently breaks the wizard, the admin panel and the backup rotation alike — docs/admin-panel.md already said so and the README contradicted it. And nothing documented that content/install.json holds the Immich API key, so a backup of content/ is a backup of credentials.
Also adds a Contributors section crediting the people whose work is in the project.
* feat(admin): floating save bar pinned to the bottom of the viewport (#426)
The save bar was the first, static element inside .admin-main. Since the
document itself scrolls, it left the viewport as soon as you scrolled down
— editing a subpage section or a settings block far below gave no visible
dirty hint and no reachable save button.
Replace it with a bottom-pinned floating bar shared by both editors. It
only mounts while there is something to act on (unsaved changes, an
in-flight save, or a result message), so it never covers content while
idle. The success message clears after 5s, which hides the bar again.
- new SaveBar component, replacing the duplicated markup in PageBuilder
and SettingsEditor
- adopt the previously unused .floating-save-bar CSS: slide-in keyframes
(carrying translateX(-50%)), dirty border, wrap/gap, mobile variant and
a prefers-reduced-motion opt-out; z-index 900 keeps it below modals
- .admin-main gains bottom padding so the bar never covers page content
- drop the now-unreachable idle hints: the ⌘S hint becomes a button title,
the Live Sync note becomes a static line above the settings layout
- header "Site" link now bypasses the cache (/?fresh=1), so the preview
stays reachable when the bar is hidden
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(admin): themed listbox for the album sort control (#427)
The per-album "Photo order" control was a native <select>. Because an
<option> cannot contain markup, its choices could not use the admin
panel's currentColor SVG set — the reason #415 had to fall back to plain
text after dropping the emoji labels.
Adds a Listbox component implementing the ARIA combobox/listbox pattern,
so options can render a glyph beside their label:
- DOM focus stays on the trigger button; the active option is announced
through aria-activedescendant, keeping focus management to one element
- Keyboard: arrows (clamped at the ends, like a native select),
Home/End, Enter/Space to commit, Escape to cancel, Tab commits and
moves on, and prefix type-ahead with a repeat-key cycle
- Outside dismissal on pointerdown, so a press starting outside closes
the popup before the click retargets
The sort options now carry an icon each: calendar, arrow-down, arrow-up,
an A→Z marker, and the drag glyph for manual — matching the badge the
album card already shows. SORT_LABELS stays plain text because the badge
feeds it to a title attribute.
Closes #417
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore(git): ignore generated design handoff folders
* fix(admin): stop the album sort dropdown clipping, and simplify the panel (#428)
* fix(admin): render listbox popup in a portal so it escapes modal clipping
The sort control's popup was position:absolute inside three clipping ancestors
(.album-drawer overflow:hidden > .album-drawer-body overflow-y:auto >
.modal-right-column overflow-y:auto), so it was cut off at the bottom of the
Edit Album Details modal. z-index could not help: clipping happens before
stacking. It also always hung below the trigger, with no flip.
Portal the popup to document.body and position it from the trigger's viewport
rect, flipping above when there is more room there and capping max-height to the
space available so it scrolls internally rather than overflowing. It is
dismissed on scroll and resize, since the anchor is only measured on open.
The outside-pointerdown handler now also treats the popup as inside; without
that the portalled list counts as outside, and a click on an option would close
it before the selection commits.
* refactor(admin): single-column album details panel
The left column spent 1.2fr on a cover thumbnail and three stat rows, two of
which (Status, Custom Hero) only restated the form fields opposite them, while
the form itself was squeezed into 1.8fr behind a second nested scroller.
Stack the drawer in one column instead: cover as a banner, then the form at full
width, with the UUID moved into the footer. The body is now the only scroll
container, which also removes one of the ancestors that was clipping the sort
popup.
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* fix(admin): give the portalled listbox its theme back, lock the page behind modals (#429)
Two defects in the album editor.
The sort listbox popup painted transparent. Every --admin-* variable is
scoped to .admin-layout, and since the popup moved into document.body it
sits outside that subtree, so background/border/text all resolved to
nothing. It now carries its own copy of the palette in both colour
schemes, plus the font it no longer inherits.
Scrolling inside a modal also moved the page behind it. No admin modal
ever locked body scroll: the overlay covers the page but does not absorb
wheel events, so a scroll starting on the backdrop — or continuing past
the end of the modal's own scroller — reaches the document.
overscroll-behavior cannot cover that case on its own, because it only
applies to scroll containers and the backdrop is not one.
useScrollLock freezes body while a modal is open and compensates for the
width the scrollbar gives up, so nothing shifts underneath. Each lock
restores the inline value it found, which makes nesting work without a
counter — a picker opened from the drawer restores the drawer's hidden
state instead of clearing it. Applied to both drawers and all four
picker modals, which share the defect. overscroll-behavior: contain on
the scrollers stops the chaining they can handle themselves.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
* fix(security): allow unsafe-eval in CSP script-src during dev mode
* feat(journal): add Journal Studio with split-screen editor, live preview and dedicated routes
* fix(build): separate client parser from server journal-service and fix build type errors
* fix(journal): Fotos rendern, Layout und Lightbox-Steuerung korrigieren
Fünf unabhängige Fehler in der öffentlichen Journal- und Essay-Ansicht:
* Foto-Blöcke wurden nie angezeigt. EssayView löst ein Foto über
PhotoItem.id auf, und das sind verschlüsselte Tokens — die Blöcke aus
dem Markdown tragen aber rohe Asset-UUIDs. Jede Zuordnung lief ins
Leere, und nicht auflösbare Blöcke werden still verworfen. Die IDs
werden jetzt serverseitig übersetzt; damit verschwinden nebenbei die
rohen UUIDs aus dem Client-Payload, die dort laut Projektregel nie
auftauchen dürfen. Nicht auflösbare Referenzen werden geloggt.
* Überschriften standen weiter links als der Fließtext. Jeder Block
hatte einzeln `max-width: 68ch`, und `ch` ist relativ zur Schrift des
Elements — die Serif-Überschrift ergab eine breitere Spalte. Ersetzt
durch eine gemeinsame, font-unabhängige `--essay-measure`.
* Fullbleed-Fotos bluteten nur nach links aus: `.essay-figure` ist
`width: 100%`, wodurch die Box überbestimmt ist und der negative
rechte Margin verworfen wird.
* Foto-Paare stellten beide Bilder gleich breit dar, wodurch ein
Hochformat neben einem Querformat doppelt so hoch wurde. Jetzt eine
Justified-Reihe: Breite proportional zum Seitenverhältnis, dadurch
exakt gleiche Höhen ohne Beschnitt.
* Die Lightbox hatte im Journal keine Tastatursteuerung und keine
Scroll-Sperre — beides lag in PhotoGrid, also nur im Album-Raster.
Verlagert in die Lightbox selbst, der doppelte Handler in PhotoGrid
entfernt (sonst wäre pro Pfeildruck zwei Bilder weitergesprungen).
Zusätzlich sind Proofing-Herzen und das EXIF-Panel im redaktionellen
Modus abschaltbar und im Journal aus: Client-Proofing ist ein
Übergabe-Workflow für Alben, keine Funktion einer Reportage. Auf der
Übersicht wurde der Untertitel doppelt gerendert, weil extractExcerpt()
auf ihn zurückfällt.
* feat(admin): eigene URLs für alle Bereiche und überarbeiteter Journal Studio
Das Panel war eine einzige Client-Komponente, die ihre Tabs über
useState umschaltet: keine Route pro Bereich, kein Deep-Link, kein
Zurück-Button, und nach jedem Reload landete man wieder auf "Pages".
Auch der Journal-Editor hatte keine eigene URL.
Jetzt sieben echte Routen (/admin, /admin/pages, /admin/journal,
/admin/journal/<slug>, /admin/settings, /admin/settings/<section>,
/admin/analytics). Die Shell mit Auth-Prüfung und Kopfzeile sitzt im
Layout und umschließt children — so bleibt die Session-Prüfung über
Routenwechsel montiert, statt bei jedem Tab-Wechsel neu zu laufen und
den Spinner aufblitzen zu lassen. Unbekannte Settings-Sections fallen
auf General zurück.
Im selben Zug der Journal Studio, der die Routing-Änderung ohnehin
berührt:
* Alle Eingabefelder nutzten die Klasse `admin-input`, die in keinem
CSS definiert war — sie fielen auf Browser-Defaults zurück. Jetzt
gestylt, gescoped, damit der generische Name nicht ins übrige Panel
leckt.
* Der Split View ist per Divider verschiebbar (Maus, Tastatur,
Doppelklick zum Zurücksetzen), Breite in localStorage.
* Das Studio bricht aus der 1120px-Spalte aus, die Preview hat Platz.
* Die Preview zeigte für jedes Foto fest 3:2 und log damit über das
Layout der veröffentlichten Seite; die echten Verhältnisse werden
jetzt an den Thumbnails gemessen.
* Blöcke mit alten positionsbasierten Foto-Referenzen werden markiert —
die lassen sich auf einer eigenständigen Journal-Seite nicht auflösen
und verschwanden bisher stillschweigend.
* Blocktyp-Chips als gemeinsame Komponente, monochrom und per Icon
unterschieden statt in einer eigenen Blau/Lila/Orange-Palette, die
sonst nirgends im Panel vorkommt.
* Emojis raus, SVG-Icons rein.
* fix(admin): Designsystem-Lücken im Panel schließen
Eine systematische Suche nach Klassen, die im TSX benutzt, aber in
keinem CSS definiert sind, hat vier Stellen mit demselben Muster
gefunden:
* `admin-btn-secondary` und `admin-btn-danger` existierten nicht.
Sämtliche Löschen-Buttons sahen aus wie normale Buttons — keine
Warnwirkung. Als `.admin-btn.admin-btn-danger` geschrieben, weil die
Varianten mit den Größenklassen kombiniert werden, die weiter unten
ihren eigenen Hintergrund setzen.
* `page-progress-fill` fehlte: die Balken im Analytics-Panel waren
leere Rinnen, der Wert unsichtbar. Ebenso `analytics-empty` und
`admin-loading-container`.
* `subpage-drawer-section`, `subpage-drawer-title-group` und
`input-with-icon` im PageBuilder waren wirkungslos und teils mit
Inline-Styles notdürftig kompensiert.
* `grid-card-preview` fehlte, wodurch alle Layout-Vorschauen auf Höhe
null kollabierten. Nur "Showcase" war sichtbar — und das aus Versehen,
weil sein Tile `hero-tile` hieß und damit die `aspect-ratio` einer
fremden PageBuilder-Klasse erbte. Umbenannt.
Dazu die Theme-Vorschauen, die praktisch identisch aussahen: alle
sieben zeigten dieselben drei quadratischen Kacheln, einziger
Unterschied war die Hintergrundfarbe. Sie spiegeln jetzt die echten
Preset-Daten aus lib/config/theme.ts — Schriftcharakter, Eckradius,
Rasterdichte, Passepartout bzw. Schatten.
Die Kopfzeile richtet sich an einer gemeinsamen Steuerhöhe aus. Der
"Site"-Link war höher als die Buttons daneben: `.admin-btn` hatte keine
explizite line-height, und <button> bekommt box-sizing: border-box vom
Browser, <a> nicht. Die Statusanzeige setzte weder Schriftfamilie noch
-größe und fiel auf den Browser-Default zurück — sie war nie Teil des
Designsystems. Jetzt gleiche Höhe und Typografie, Pillenform als
Zustandsanzeige, Trenner zur Aktionsgruppe. Das dauerhafte Pulsieren
des grünen Punkts entfällt; Animation bleibt den Zuständen vorbehalten,
die Aufmerksamkeit brauchen.
Restliche Emojis durch SVG-Icons ersetzt.
* feat(admin): add scroll-to-top toggle and SVG arrow
The back-to-top button was always on and drew its arrow with the ↑ text
character. Add a `scrollToTop` setting (General section in the admin
panel, defaults to on so existing sites are unaffected) and replace the
glyph with a stroked SVG that inherits currentColor across all presets.
Scroll handling is now rAF-throttled, and prefers-reduced-motion turns
off both the hover lift and the smooth scroll.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* security(journal): escape author HTML instead of filtering it
CodeQL flags 26 high-severity alerts on the release PR; the bulk of them
point at sanitizeHtml(), a denylist of three regexes whose output goes
into dangerouslySetInnerHTML. I reproduced the bypasses before changing
anything — all four pass the old filter:
<img src=x onerror=alert(1)> unquoted handler, regex wants quotes
<img src=x onerror='alert(1)'> single quotes, same reason
<svg onload=alert(1)> not a <script> tag
javasjavascript:cript:alert(1) one-pass replace, the rest recombines
Plus an attribute breakout in the link renderer, which interpolated the
URL into href="${url}" unescaped:
[x](" onmouseover=alert(1) x=") -> <a href="" onmouseover=alert(1" ...>
Replaced with escaping. Every `&<>"'` in author text becomes an entity
before any markdown rule runs, so the only tags in the output are the
ones renderInlineMarkdown builds itself and a URL cannot leave its
attribute. Link schemes are now an allowlist (http, https, mailto, tel,
relative, anchor); anything else renders as plain text.
The serializer needed decodeHtmlEntities to match: without it a save
would write `<` into the Markdown file and the next parse would
escape the `&` again, corrupting the text further on every round trip.
A test covers parse -> serialize -> parse stability.
Two existing tests asserted the old behaviour and were rewritten rather
than deleted: one expected the script tag to be *removed* from the
output, the other expected the substring "onerror" to be absent. Both
now assert the property that matters — that no tag can form — because
`onerror` surviving as inert text is fine.
Exploitation required admin access either way: journal and essay text
comes from the studio or from content/*.md. This removes the weakness,
it does not close an open door.
* fix(journal): contain file paths, fix frontmatter escaping, defuse two regexes
Follow-up to the CodeQL findings on release PR #432. I went through all 23
alerts individually; this addresses the ones that are worth addressing and
documents the ones that are not.
Path containment (12 alerts, all false positives). Every path built from a
slug goes through isValidSlug(), a strict `/^[a-zA-Z0-9_-]+$/` allowlist —
verified against `../../etc/passwd`, `foo/../../bar`, `a/b`, `..` and
friends, all rejected. CodeQL cannot see it because the check lives in
another module. Added a containedPath() helper that resolves the path and
refuses anything landing outside the content directory: a second lock on a
door that was already locked, and a barrier the analysis can follow. Also
removed the dead `slug.endsWith('.md')` branches — isValidSlug rejects
dots, so they could never be taken.
Frontmatter escaping (1 alert, a real bug). The serializer escaped `"` but
not `\`, and the parser never unescaped anything. A title containing either
character gained a backslash on every save/load cycle:
Path C:\Users "quoted" -> Path C:\\Users \"quoted\" -> ...
Escaping now covers the backslash first and the parser undoes both. A round
trip test covers quotes, backslashes, colons and apostrophes.
Two of three ReDoS regexes (2 alerts). The heading pattern had `\s+` next to
`(.+)`, which overlap, so the engine tried every split of the whitespace run
before failing; it is `[ \t]+` and `(\S.*)` now. The quote author pattern
wrapped a lazy group around an optional group — the quadratic case — and is
an index split on the first separator now.
The third regex, the bold/italic pass, is deliberately left alone. Rewriting
it with a negated class removes the backtracking but breaks nesting:
`**bold *and italic* here**` renders as `*<em>bold </em>and italic<em> here</em>*`.
The input is Markdown from an authenticated admin, never a visitor, so the
worst case is an author slowing down their own page. The comment in the code
says so.
Not addressed: 5 js/insufficient-password-hash alerts. They ask for a slow
KDF where the code uses SHA-256/HMAC — correct in general, but the admin
password comes from an environment variable and is never stored, so a
stronger KDF protects nothing that is not already in plaintext in the
environment.
399 unit tests pass (18 new), build and tsc clean.
* fix(journal): split the quote author by scanning, not by pattern
Cleaning up after myself. #439 replaced the quote author regex
`/^(.*?)(?:\s+--\s+(.+))?$/` — a lazy group around an optional one, the
quadratic case — with `/\s+--\s+/`, and CodeQL promptly flagged the
replacement for the same reason. One reported pattern traded for another.
splitQuoteAuthor() scans for the first `--` that has whitespace on both
sides. Linear, and it needs no reasoning about the regex engine at all.
Behaviour is identical to the original regex, checked case by case
against it: separator, no separator, several separators, tabs, wide gaps,
`a--b` without spaces, leading and trailing separators, empty string. All
nine are covered by tests, plus one asserting it stays fast on a 50k
whitespace run.
* security(admin): derive the session signing key with scrypt
getSigningKey() ran the admin password through a single SHA-256 digest.
It is key derivation rather than password storage, so it was never a
stored-hash problem — but if the derived key ever leaked, one fast hash
makes recovering the password by brute force cheap. CodeQL flags it as
js/insufficient-password-hash and, for this particular alert, has a
point.
scrypt now. Two details that matter:
The salt has to be stable. A random salt would produce a different key on
every call and invalidate every session; it is derived from AUTH_SECRET,
which is deployment-specific. Verified across separate processes: same
inputs produce a verifying token, so a restart or redeploy does not log
everyone out.
The result is cached for the process lifetime. scrypt is deliberately
slow and getSigningKey() runs on every admin request — deriving per call
would add ~24ms to each one. The cache key is a digest of secret plus
password, so a rotation misses the cache and derives again, and no extra
plaintext copy of the password is kept around.
Measured: first derivation 24ms, subsequent calls 0.12ms. Password
change and secret change both invalidate an existing token, checked in
separate processes.
Note the one-time effect: existing admin sessions are invalidated by the
change in derivation, so everyone logs in once more.
* security(admin): key the signing-key cache on its inputs, not a digest
#441 removed the SHA-256 digest of the admin password from the key
derivation and introduced a new one three lines above it, as the cache
key. Same rule, same file: CodeQL reported it immediately.
The comment justifying it was wrong too. It claimed no second plaintext
copy of the password was retained, while building the digest input
allocates exactly that in order to hash it.
The cache now holds the two input strings and compares them directly.
They are the same objects the env and install caches already hold, so
nothing extra is retained, and no fast hash of the password exists in
memory for the life of the process.
Re-verified after the change, in separate processes: a token from one
process verifies in another with the same inputs, a changed password or
secret rejects it, first derivation 24ms, cached calls 0.001ms.
That leaves scrypt as the only place the password is hashed here. The
two HMACs further down operate on the derived key and on token data, not
on the password; the constant-time comparison in verifyAdminPassword is
the previously dismissed one.
---------
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: lancetm714 <108419425+lancetm714@users.noreply.github.com>
2026-08-14 16:55:07 +02:00
|
|
|
|
INSTALL_CONTENT_DIR?: string;
|
2026-03-08 21:12:52 +01:00
|
|
|
|
}
|
2026-03-07 08:18:23 +01:00
|
|
|
|
|
2026-03-08 21:12:52 +01:00
|
|
|
|
function parseEnv(): Env {
|
|
|
|
|
|
const errors: string[] = [];
|
|
|
|
|
|
|
|
|
|
|
|
const urlRaw = process.env.IMMICH_API_URL;
|
|
|
|
|
|
let apiUrl = '';
|
2026-03-08 21:38:44 +01:00
|
|
|
|
if (urlRaw) {
|
2026-03-08 21:12:52 +01:00
|
|
|
|
try {
|
|
|
|
|
|
apiUrl = new URL(urlRaw).toString().replace(/\/+$/, '');
|
|
|
|
|
|
} catch {
|
2026-03-08 21:38:44 +01:00
|
|
|
|
console.warn('⚠️ IMMICH_API_URL is invalid, falling back to empty string for setup.');
|
2026-03-08 21:12:52 +01:00
|
|
|
|
}
|
|
|
|
|
|
}
|
2026-02-14 11:47:26 +01:00
|
|
|
|
|
2026-03-08 21:38:44 +01:00
|
|
|
|
const apiKey = process.env.IMMICH_API_KEY || '';
|
2026-03-08 21:12:52 +01:00
|
|
|
|
|
|
|
|
|
|
const cacheTtlStr = process.env.CACHE_TTL;
|
2026-03-12 11:30:45 +00:00
|
|
|
|
const cacheTtl =
|
|
|
|
|
|
cacheTtlStr && !isNaN(parseInt(cacheTtlStr, 10)) ? parseInt(cacheTtlStr, 10) : 300;
|
2026-03-08 21:12:52 +01:00
|
|
|
|
|
Close the gaps left after v0.9.2's outage handling: route tests, stale cache, loading states (#397)
* docs: brainstorm resilience/adoption features and plan follow-up work
Brainstorm and deep-dive on where Immich Folio breaks under Immich
outages and what blocks self-hosted adoption, plus an implementation
plan for the remaining gaps after v0.9.2 (route-handler tests,
stale-while-error caching, loading skeletons).
* test: enable route-handler tests and cover admin guards and image content-type
vitest.config.ts only ran lib/__tests__/**, so route handlers were
structurally untestable. Two mechanisms live only at the route level and
cannot be extracted the way lib/imageSize.ts was: the admin auth guards
and the image proxy's Content-Type sanitisation (the stored-XSS fix from
c2fa8e7). Both were correct but unverified by any test.
Adds a table-driven test over all ten guarded /api/admin/* handlers,
including a filesystem walk that fails if a future route is added
without a row in the table. Adds a route-level test for
/api/image/[id] covering token validation, the six Content-Type
rewrite cases, ETag/304, and the 404 vs 503 (ImmichUnavailableError)
distinction.
Verified both catch real regressions by temporarily removing the admin
guards and the Content-Type rewrite and confirming the expected tests
fail, then restoring and confirming green again.
134 new tests (244 total, up from 210), tsc 0 errors, build succeeds.
* feat: serve stale cache entries while Immich is unavailable
v0.9.2 (#394) correctly split "gone" (404/410) from "unavailable"
(ImmichUnavailableError), and pages now render a 5xx instead of a false
404 during an outage — but that 5xx still replaced a gallery that had
loaded fine seconds earlier. lib/cache.ts deleted expired entries on
access, so nothing survived long enough to fall back to.
Cache entries now carry two deadlines: staleAt (the existing CACHE_TTL,
after which get() still reports a miss) and hardExpiresAt (the new
STALE_MAX_AGE, default 24h, after which the entry is dropped entirely).
Between the two, getStale() can still return the data.
getAlbums(), getAlbum() and getAssetInfo() catch ImmichUnavailableError
and fall back to getStale() before propagating. The two MISSING-sentinel
writes (definitive 404s) are deliberately left on the short cacheTtl
window — a corrected asset ID should take effect after the normal TTL,
not sit in the 24h stale window meant for outages.
STALE_MAX_AGE=0 disables the fallback. Verified the fallback only
engages on a real outage, not on a cache hit at the TTL boundary, and
that a definitive 404 still reports missing rather than falling back.
256 tests (up from 244), tsc 0 errors, build succeeds.
* feat: add loading skeletons for dynamic gallery routes
Every gallery route is force-dynamic and waits on a live Immich
round-trip — up to IMMICH_TIMEOUT_MS when the server is struggling —
with nothing rendered until it resolves. loading.tsx renders instantly
while the page component awaits its data.
Not browser-verified against live data: this worktree has no
.env.local and only .example content files, so the app boots into
setup mode (confirmed the dev server starts cleanly and the setup
screen renders, HTTP 200). tsc 0 errors, build succeeds, no new lint
errors.
* docs: document stale cache window, image token revocation, route-test convention
- README: STALE_MAX_AGE env var and a "Behaviour when Immich is
unreachable" section (stale-serves-until-STALE_MAX_AGE, 503 not 404,
outages never cached).
- docs/gallery-config.md: new "Image Token Revocation" subsection under
System & Security — un-publishing an album does not revoke already
issued image URLs; rotating AUTH_SECRET does.
- CLAUDE.md: updated cache description to cover the stale window and
MISSING sentinel, and a new code-convention note that route-level
logic (guards, header sanitisation) should get a route-level test
rather than being extracted into lib/ just to reach the test runner.
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
2026-08-07 11:07:34 +02:00
|
|
|
|
// How long an expired cache entry may still be served while Immich is
|
|
|
|
|
|
// unreachable. Keeps the public gallery online across an Immich restart
|
|
|
|
|
|
// instead of showing an error page. 0 disables the fallback.
|
|
|
|
|
|
const staleMaxAgeStr = process.env.STALE_MAX_AGE;
|
|
|
|
|
|
const staleMaxAge =
|
|
|
|
|
|
staleMaxAgeStr && !isNaN(parseInt(staleMaxAgeStr, 10)) ? parseInt(staleMaxAgeStr, 10) : 86400; // 24 hours
|
|
|
|
|
|
|
fix: Immich outage handling, image tiers, admin save safety (#394)
* fix: migrate deprecated middleware convention to proxy
Next.js 16 deprecated the `middleware` file convention in favour of
`proxy`. The build warned on every run, and the convention is slated for
removal in a future major — at which point Next.js would silently stop
invoking this file. That is not a cosmetic failure: proxy.ts owns the
Content-Security-Policy and the per-request nonce, so an uninvoked file
means every document response ships with no CSP at all, with nothing in
the build output to catch it.
Renames middleware.ts to proxy.ts and the exported middleware() to
proxy(), per the official migration. The config.matcher export is
unchanged, so route coverage is identical.
Adds lib/__tests__/proxy.test.ts, which pins the export name and
verifies the behaviour a silent regression would take out: a CSP is set,
the nonce in the policy matches the one forwarded to pages via x-nonce,
nonces differ per request, x-pathname still reaches the root layout, and
the matcher continues to cover /admin.
Verified: tsc 0 errors, 105 unit tests passing (98 before), production
build succeeds with 0 deprecation warnings.
* fix: stop reporting Immich outages as missing albums
lib/immich.ts request() returned null for every failure mode: a 5xx, a
rejected API key, a network error, and a non-JSON body all looked exactly
like "Immich answered and said this does not exist". Callers had no way
to tell the two apart, and app/[...path]/page.tsx calls notFound() on a
null album in five places.
The result was that while Immich was down or returning 5xx, every album
URL served a hard 404 — for albums that exist. Visitors saw "not found",
and crawlers were free to de-index real content over a transient outage.
request() now throws ImmichUnavailableError for anything that is not a
definitive answer about the resource, and reserves null for 404 and 410.
Pages let that propagate, so an outage surfaces as a 500 rather than a
404. Two callers must not propagate and are handled explicitly:
- ping() catches and returns false — it answers "is Immich reachable",
and both callers (/api/health, /api/admin/status) render that as a
status field.
- /api/exif and /api/map return 503 with Retry-After instead of 404 and
an empty marker list respectively. An empty map is otherwise
indistinguishable from a gallery with no GPS data.
Adds 10 tests: each failure mode maps to the right outcome, getAlbum and
getAlbums propagate rather than reporting empty, ping stays non-throwing,
and a failed request leaves no stale entry in the in-flight coalescing
map — which would otherwise replay the failure after Immich recovered.
Verified: tsc 0 errors, 115 unit tests passing (105 before), production
build exit 0, eslint 0 errors on all touched files.
* fix: bound how long Immich requests may hang
lib/immich.ts set no timeout on any of its three fetch call sites, so the
only ceiling was undici's built-in default, measured in minutes. Every
page is force-dynamic, which means an Immich that accepts connections but
never answers held each visitor's render open for that whole window —
and since nothing caches the failure, load kept arriving.
Adds IMMICH_TIMEOUT_MS (default 15000, floored at 1000 so a stray 0
cannot mean "no timeout"), threaded through env.ts and AppConfig like
CACHE_TTL.
The two kinds of call need different treatment:
- request() carries a whole-exchange AbortSignal.timeout(). JSON
payloads are small, so one budget can cover headers and body.
- streamAsset()/streamVideo() use an AbortController whose timer is
cleared in a finally the moment response headers arrive. A large
original photo or a long video is legitimately slow to transfer; a
whole-request timeout would truncate playback and break seeking, since
every range request would restart the clock.
A timed-out request surfaces as ImmichUnavailableError naming the budget,
so a misconfigured value is diagnosable from the log line alone.
Adds 5 tests, including the two that pin the distinction: a stream whose
headers have arrived is not aborted an hour later, while a stream that
never sends headers still is.
Verified: tsc 0 errors, 120 unit tests passing (115 before), production
build exit 0, no new eslint errors (the one error in env.ts is
pre-existing — confirmed against a clean tree).
* fix: stop caching Immich outages as permanently missing media
f3 split "gone" from "unavailable" in request(), but streamAsset() and
streamVideo() have their own fetch calls and were left returning null for
every failure. Both proxy routes turn null into a 404.
That 404 carried no Cache-Control, and a 404 is heuristically cacheable
per RFC 9111 — on URLs the success path serves as `immutable`. So a
transient outage, or a timeout introduced in 370e668, could pin a broken
image or an unplayable video in the visitor's browser cache well after
Immich recovered.
Both stream methods now throw ImmichUnavailableError for transport
failures, timeouts, and any error status other than 404/410, and reserve
null for genuinely absent media. A 200 with no body counts as
unavailable, not as missing. The routes translate that into 503 with
Retry-After, and both error responses — the 503 and the real 404 — now
send Cache-Control: no-store.
streamVideo keeps accepting 206, so range requests and seeking are
unaffected.
Adds 7 tests covering each failure mode for both methods plus the 206
path, and updates the F4 timeout test, which asserted the old null return.
Verified: tsc 0 errors, 127 unit tests passing (120 before), production
build exit 0, eslint 0 errors on touched files.
* perf: cache "does not exist" answers from Immich
getAlbum() and getAssetInfo() cached successes but discarded a definitive
"not found", so a missing resource was re-queried upstream on every call.
Measured before: 5 sequential lookups of a missing asset produced 5
upstream fetches. After: 1.
This is not hypothetical. app/page.tsx resolves every gallery.yaml hero
ID through getAssetInfo() on each render, and the homepage is
force-dynamic — so a hero photo deleted from Immich but still listed in
gallery.yaml cost one 404 round-trip per page view, indefinitely.
Absence is stored as a frozen MISSING sentinel because the cache cannot
distinguish a miss from a stored null; both read back as null.
Only definitive 404/410 answers are cached. An outage throws
ImmichUnavailableError and is deliberately never stored — caching one
would keep the gallery broken after Immich recovered. A test pins that:
a 503 followed by a healthy response must reach Immich the second time.
Stored under the normal cacheTtl rather than a separate shorter negative
TTL, because the reason for a shorter one no longer applies: outages are
not cached at all now, and both invalidation paths (the Immich webhook
and the admin save/reload) clear the entry immediately, so a corrected
asset ID takes effect at once rather than waiting out a TTL.
Also raises the test config's cacheTtl from 0, which expired entries in
the same millisecond they were written and would have made any caching
assertion flaky. Test isolation comes from cache.clear() in beforeEach.
Verified: tsc 0 errors, 131 unit tests passing (127 before), production
build exit 0, eslint 0 errors on touched files.
* fix: add themed error and not-found pages
The app had no app/error.tsx, app/not-found.tsx or global-error.tsx, so
every failure fell through to Next.js's unstyled default page.
That gap became visible with 253ade0, which made an Immich outage produce
a 500 instead of a false 404 — correct, but what a visitor saw was a bare
framework error screen with none of the site's theming.
The two pages also carry the distinction the error split was for, and
they carry it structurally rather than by inspecting the error: notFound()
renders not-found.tsx, a throw renders error.tsx. That matters because
Next.js strips error messages in production and replaces them with a
digest — this component genuinely cannot tell an Immich outage from any
other server error, so it does not claim to. It frames the failure as
temporary and offers reset(); not-found.tsx deliberately does not, since
a real 404 is not transient.
The digest is shown so a user report can be traced to a log line. The raw
error message is never rendered — it can carry an upstream URL, a header
or a config value.
Both pages reuse .empty-state, which existed in globals.css but was
unreferenced anywhere in the app; extended with title, actions and button
elements matching the PasswordGate button, the only other action button
in the codebase.
Tested by server-rendering both components (react-dom/server, no new
dependencies) — 8 tests covering the styling hook, the retry and home
affordances, digest handling, and that a message containing an API key or
host address does not reach the markup. Note vitest.config.ts only
includes lib/__tests__/**/*.test.ts, so the test uses createElement
rather than JSX; a .tsx file there would never run.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the app boots into setup mode and neither
Playwright nor a dev server can exercise these routes here.
Verified: tsc 0 errors, 139 unit tests passing (131 before), production
build exit 0 with /_not-found registered, eslint 0 errors.
* perf: drop the unused originalUrl from every photo item
PhotoItem declared originalUrl and toPhotoItems computed it for every
asset, but a repo-wide grep found exactly two references: the field
declaration and the assignment. Nothing ever read it — the grid uses
thumbUrl, the lightbox uses previewUrl.
The cost was not the encryption. PhotoItem is a server-to-client prop, so
the dead field was serialized into the RSC payload and shipped to the
browser for every photo.
Measured: 137 bytes per asset once serialized, so 6.7 KB for a 50-photo
album and 66.9 KB for a 500-photo one — transferred and parsed on every
page view, never read.
Removed rather than wired up to a "view original" affordance. Exposing
full-size originals is a bandwidth and privacy decision for the project
to make deliberately, not something to infer from a dangling field.
No unit test: this is a deletion whose correctness is exactly "nothing
referenced it", which tsc and a repo-wide grep already establish. A test
asserting the absence of a field would restate the type checker.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint unchanged at 4 pre-existing errors on the touched files (unused
imports, confirmed identical against a stashed clean tree).
* chore: remove four orphaned imports from the catch-all route
Link, Image, PhotoGrid and BackLink were imported by app/[...path]/page.tsx
and never used — 4 of the 6 no-unused-vars errors in the repo baseline.
Verified they are leftovers, not a dropped render path: PhotoGrid and
BackLink are both rendered by AlbumDetailView, which imports them itself,
and BackLink is also used by the impressum and map pages. Rendering was
extracted out of page.tsx into AlbumDetailView/SubpageGridView at some
point and the imports stayed behind. `type PhotoItem` is still needed by
toPhotoItems, so that import narrows to a type-only import rather than
disappearing.
No bundle win, measured: .next/static is 1084 KB across 20 JS chunks both
before and after, byte-identical. Turbopack already tree-shook them. The
backlog note that predicted unused client-component imports would hold
them in the client graph was wrong, and has been corrected there.
The value is readability and lint: repo errors drop from 54 to 50, and
page.tsx no longer reads as though it renders the grid itself.
chore: rather than fix:/perf: — nothing behaves differently and the
measurement showed no performance change, so either of those labels would
misdescribe it. CLAUDE.md lists chore: as a valid prefix.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint 0 errors on the touched file.
* fix: make ?w= able to narrow the image tier instead of being ignored
The image proxy resolved its Immich rendition with `?size=` winning
whenever present, falling back to `?w=` otherwise. lib/urls.ts always
writes `?size=`, so for every URL this app generates the width branch was
unreachable and widthToSize() was dead code — while CLAUDE.md documented
the width mapping as the actual behaviour.
`?size=` is now a ceiling rather than an override: when both are present
the smaller tier wins, so a width can narrow the request but never widen
it. That asymmetry is the point. Letting width win outright — the obvious
reading of the docs — would have been a real regression: next/image emits
widths up to 3840 and widthToSize(1920) is 'original', so every large
display would have started downloading full-size originals in place of
previews. A test pins that specific case.
No measured change today. The only URLs carrying both parameters are
?size=preview with next/image widths, whose smallest vw-based deviceSize
is 640 — already preview either way. This is a correctness fix that makes
the documented contract true and testable, not a performance win.
Moved the logic to lib/imageSize.ts so it can be tested at all:
vitest.config.ts only collects lib/__tests__, so nothing inside a route
file is reachable from the suite. 9 tests cover the tier boundaries, both
single-parameter paths, malformed input, and the ceiling behaviour.
CLAUDE.md updated to describe the ceiling rule rather than a plain
width mapping.
Verified: tsc 0 errors, 148 unit tests passing (139 before), production
build exit 0, eslint 0 errors on touched files.
* fix: validate the per-album heroImage UUID from gallery.yaml
Every asset ID in gallery.yaml is routed through validateUuid before it
can reach encodeAssetId, which encrypts whatever it is handed. One was
not: for a per-album entry the album ID (the map key) was validated while
the heroImage value was stored raw.
A typo there produced a perfectly well-formed image URL that 404s forever,
with nothing in the log naming the cause — the failure surfaced as a
broken hero image on an album page, far from the line that caused it.
This is narrower than the finding that prompted it, which proposed
guarding encodeAssetId itself. Validation belongs at config load, where it
runs once and can name the file and the album; a guard in encodeAssetId
would run on every render and duplicate validateUuid at the wrong layer.
The heroImage path was the only unvalidated route into it.
validateUuid warns and substitutes the nil UUID rather than throwing, so a
bad ID still degrades to a missing image instead of taking the page down —
and since f511e62 that 404 is negative-cached, so it costs one upstream
request rather than one per render.
Adds 4 tests. The two that pin the wiring were mutation-checked: both fail
against the unfixed code and pass with it.
Verified: tsc 0 errors, 152 unit tests passing (148 before), production
build exit 0, eslint 0 errors on touched files.
* fix: keep /admin reachable when gallery.yaml cannot be derived
getConfig() throws on three shapes the admin page builder is able to
write: a gallery with neither albums nor subpages, a subpage with no name,
and a subpage with neither albums nor sections.
RootLayout called it unguarded, and a throw in a *layout* is not caught by
app/error.tsx — that needs a global-error boundary, which this app does
not have. /admin renders inside that same layout, so one bad save took
down the site together with the only tool that could undo it. Recovery
meant editing content/gallery.yaml on the host, or exec'ing into the
container. This is issue #326's failure mode reached through an invalid
config instead of a missing one.
Adds getConfigOrNull(), which returns null rather than throwing and logs
the underlying error plus the path to content/.backups/. The layout treats
null exactly like needsSetup: setup screen for visitors, children rendered
for admin paths so the panel still loads. generateMetadata() is guarded
too — it also calls getConfig(), and it also runs before any error
boundary.
This is the half of the finding that makes every broken config
recoverable, including ones this app never wrote: a hand-edited file, a
restored backup, a config from an older version. The other half —
validating in the admin PUT so the bad state is never written — is filed
separately; it prevents new breakage but cannot repair existing breakage,
so it was not the one to do first.
Adds 10 tests. Three pin the premise by asserting getConfig() really does
throw on each shape, so the fallback cannot quietly become dead weight.
One asserts the needsSetup path still returns a config rather than null,
since missing and invalid are different states that must not collapse.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the layout/error-boundary interaction is
reasoned from Next.js semantics and covered at the getConfig level only.
Verified: tsc 0 errors, 162 unit tests passing (152 before), production
build exit 0, eslint 0 errors on touched files.
* fix: reject an admin save the site would not be able to load
The admin PUT validated only the shape of the hero field, then wrote. None
of getConfig()'s three throw conditions was checked, so the page builder
could persist a gallery.yaml that cannot be loaded — and the user got
"Saved successfully" for it. Since 95877af that no longer locks anyone
out, but the public site still drops to the setup screen until someone
notices.
Extracts the gallery half of getConfig() into deriveGallery(), which the
route now runs against the proposed gallery before writing. Two
alternatives were rejected on purpose:
- Re-checking the three conditions in the route would put a second copy of
the rules next to the first, and the two would drift.
- Write-then-verify-then-roll-back leaves a window in which other requests
read the broken config.
The extraction is a pure move: no logic changed inside it, and all 162
pre-existing tests pass untouched, which is what makes that claim
checkable rather than asserted.
The 400 carries the derivation's own message, so the admin UI names the
offending subpage instead of saying "invalid gallery" — a builder with a
dozen subpages needs to know which one. A test pins that the message
identifies it.
Adds 8 tests: each rejected shape, the named-subpage message, and the
valid structures (standalone albums, subpage membership, sections,
per-album overrides and hero images) so the extraction cannot silently
lose behaviour.
Verified: tsc 0 errors, 170 unit tests passing (162 before), production
build exit 0, eslint 0 errors on touched files.
* fix: hash subpage passwords off the main thread
verifyScrypt used crypto.scryptSync, measured at 22.8ms per call — 22.8ms
during which the Node process serves nothing else, because the sync
variant runs on the main thread rather than the libuv threadpool.
The /api/auth rate limit does not contain this. It allows 10 attempts per
minute, but the bucket key is the client IP, and with the default
TRUSTED_PROXY_HOPS=0 that IP is read from x-real-ip / x-forwarded-for —
both attacker-controlled. lib/rate-limit.ts already documents that those
headers are spoofable at hops=0; what followed from it was not written
down anywhere: one attacker rotating a header gets unlimited buckets, and
each request buys 22.8ms of dead process. Roughly 44 requests per second
saturates it.
Switches verifyScrypt and generateScryptHash to the async crypto.scrypt,
which runs on the threadpool. This does not stop the spoofing — that is
inherent to hops=0 and already documented — it removes the amplification
that made spoofing worth doing.
authenticate() becomes async. It had exactly one caller
(app/api/auth/route.ts). isAuthenticated() is deliberately left
synchronous: it only computes an HMAC, and it is called from server
components.
Adds 4 tests, two of which pin the fix and were mutation-checked against
the old implementation — both fail with scryptSync and pass with scrypt.
One asserts scryptSync is never called; the other asserts a 2ms interval
still fires during a hash, which is the property that actually matters. A
slower machine lengthens the hash and yields more ticks, so that test
fails on a blocked loop rather than on a busy one.
Also adds a scrypt-hashed fixture subpage, so the suite exercises
verifyScrypt rather than only the deprecated plaintext fallback.
Verified: tsc 0 errors, 174 unit tests passing (170 before), production
build exit 0, eslint 0 errors on touched files.
* fix: warn when TRUSTED_PROXY_HOPS cannot identify a client
getClientIp falls back to a shared 'unknown' bucket whenever the
X-Forwarded-For chain is shorter than the configured hop count, or the
header is absent with hops > 1.
That fallback is correct — the alternative is trusting a spoofable entry —
but it was silent, and silence is the problem. Every unidentified request
shares one bucket per endpoint, and this is an image proxy: a single
50-photo grid issues ~50 /api/image requests, so the default
RATE_LIMIT_RPM=1500 is collectively spent by roughly 30 page loads a
minute. Someone who sets TRUSTED_PROXY_HOPS=2 for Cloudflare-plus-nginx
and then reaches nginx directly gets a site that 429s everyone, with
nothing in the log pointing at the cause.
Warns once per process, naming the configured hop count and what was
actually seen — the chain length, or which header was missing. Once, not
per request: enough to diagnose without flooding the log.
The fallback itself is unchanged. Trusting a shorter chain would hand
every client its own bucket key and defeat the limiter entirely.
No warning at hops=0, where the shared bucket is the documented
consequence of not declaring a proxy rather than a misconfiguration.
Adds 6 tests covering each fall-through path, the once-per-process
behaviour, and both silence cases. The once-per-process flag needs a test
seam to re-arm, hence the __resetProxyWarningForTests export.
Verified: tsc 0 errors, 180 unit tests passing (174 before), production
build exit 0, eslint 0 errors on touched files.
* fix: send Retry-After on every rate-limited response
/api/image, /api/video and /api/exif each computed the retry delay,
interpolated it into a console.warn, and then returned a 429 without it —
the number was calculated and thrown away. /api/og did not compute it at
all. Only /api/auth and /api/map sent the header.
This matters more for an image proxy than for a typical API: one gallery
page issues around fifty /api/image requests, so a burst that trips the
limit produces fifty rejections carrying no back-off signal. The browser
and next/image then retry on their own schedule — a retry storm against a
server that just reported it was overloaded.
Adds Retry-After and Cache-Control: no-store to all four. no-store because
the success path serves these same URLs as `immutable`.
The delay now comes from a shared retryAfterSeconds() helper rather than
an inline Math.ceil in each route, which also fixes an edge every copy
shared: a window on the point of expiring yields 0, and Retry-After: 0
tells the client to retry immediately, which is the opposite of what a 429
means. A resetAt already in the past yields a negative number, which is
not valid HTTP. The helper floors at 1.
The three routes that already sent the header (/api/auth, /api/map,
/api/admin/auth) and /api/health now use the helper too — leaving four
copies of the buggy arithmetic behind while centralising the rest would
have been worse than not centralising at all.
Adds 2 tests, one of them specifically for the zero-and-negative edge.
Verified: tsc 0 errors, 182 unit tests passing (180 before), production
build exit 0, eslint 0 errors on all ten touched files.
* fix: add an escape hatch for stale browser image caches
Image and video responses are served
`Cache-Control: public, max-age=31536000, immutable`, so a browser does
not revalidate them for a year — the ETag is not even consulted. When
Immich regenerates thumbnails after a job re-run, or a photo is rotated,
the asset ID does not change, so neither does the deterministic token or
the URL. Visitors keep the old rendition, and the webhook that clears the
server-side caches cannot reach a browser. The only existing remedy was
rotating AUTH_SECRET, which invalidates every token on the site.
Corrects the framing this started from. The finding described it as an
ETag problem, but `immutable` means the ETag is never reached in the case
that matters. The URL is the only lever a browser cache responds to, so
the buster belongs there. It is also mixed into the ETag, since a matching
ETag across two different URLs would be wrong regardless.
Adds IMAGE_CACHE_VERSION, empty by default. Unset, the generated URLs are
byte-identical to before — a test asserts exactly that, because a
cache-busting feature that changes URLs when nobody asked for it would
silently discard every visitor's cache on deploy.
The value is URL-encoded at parse time; an unescaped one could otherwise
inject query parameters into every image URL on the site.
Deliberately not done: shortening max-age or dropping `immutable`. Both
trade a large, real bandwidth win for a rare correctness case, and the
cost is not measurable from this worktree. The aggressive caching stays
the default; this only adds a way out.
Adds 6 tests: default-unchanged, whitespace treated as unset, both URL
shapes, escaping, and that a bump changes the query while leaving the
token intact.
Verified: tsc 0 errors, 188 unit tests passing (182 before), production
build exit 0, eslint unchanged (the one error in env.ts is pre-existing,
confirmed against a stashed tree).
* security: validate the backup filename before restoring from it
restoreBackup() joined its argument straight into a path and copied the
result over a content file. path.join resolves `..`, so
restoreBackup('../../../../etc/passwd') would have written any readable
file over content/settings.yaml — whose contents the admin GET endpoints
hand straight back. The destination was chosen by
backupFilename.includes('gallery.yaml'), which is substring matching on
the same attacker-supplied string, so the caller picked that too.
Not exploitable today: the function has zero callers. That is precisely
why it was worth fixing rather than leaving — it is an exported function
carrying a traversal bug, waiting for whoever wires up the restore button
the admin panel does not yet have.
Hardened rather than deleted. The status endpoint already reports backup
counts, so restore is plainly intended; removing a half-built feature is
the maintainer's call, not a side effect of a security fix.
The name must now match the pattern this service itself produces, and the
destination is derived from the matched group rather than a substring
search. The redundant basename check is kept on purpose: it preserves the
guarantee if that pattern is ever loosened.
Adds 12 tests. Nine of them fail against the unguarded version —
mutation-checked by removing the guard and confirming, rather than
assuming they would.
Verified: tsc 0 errors, 200 unit tests passing (188 before), production
build exit 0, eslint 0 errors on touched files.
* fix: give each admin save its own temp file
writeYamlFile wrote to a constant `${filePath}.tmp` before renaming. The
rename is what makes the write atomic; the temp filename was not. Two
saves of the same file in flight at once shared that path, so the second
writeFile could interleave with the first's rename — publishing one save's
bytes under the other's, or a truncated mix of both. A double-clicked Save
in the page builder is enough to reach it.
The temp name now carries the pid and a per-process counter, so concurrent
writers cannot collide either within a process or across workers.
Also wraps the write and rename so a failure unlinks the temp file instead
of leaving it in content/. The cleanup swallows its own errors: failing to
tidy up must not replace the real error with a misleading one.
Adds 7 tests, and both halves were mutation-checked independently —
reverting to the fixed temp path fails two of them, removing the cleanup
fails two others. Two of the tests exist to pin properties that are easy
to regress silently: that the rename targets the temp file just written,
and that the happy path issues no unlink.
Verified: tsc 0 errors, 207 unit tests passing (200 before), production
build exit 0, eslint 0 errors on touched files.
* fix: await the stream-timeout rejection before advancing timers
`npm run test:unit` exited 1 even though all 207 tests passed. The
timeout test started `streamAsset`, advanced the fake timers past the
15s budget, and only then attached the rejection expectation. The abort
fires *during* `advanceTimersByTimeAsync`, so for one microtask turn the
promise was rejected with no handler attached — Node reports that as an
unhandled rejection, and vitest fails the whole run on it regardless of
test outcomes. CI would have blocked on a green suite.
Attaching the assertion before advancing the timers fixes the ordering
without weakening it: the same rejection and the same `captured.aborted`
check still run, just awaited after the advance rather than created
after it.
Verified: vitest exit 0, 207/207 passing, tsc 0 errors, prettier clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-05 12:04:01 +02:00
|
|
|
|
// How long to wait on Immich before giving up. Without an explicit budget the
|
|
|
|
|
|
// only ceiling is undici's built-in default (minutes), and since every page is
|
|
|
|
|
|
// force-dynamic an Immich that accepts connections but never answers would
|
|
|
|
|
|
// hold each visitor's render open for that entire window.
|
|
|
|
|
|
//
|
|
|
|
|
|
// This bounds the wait for a *response*, not the transfer of an image or video
|
|
|
|
|
|
// body — those are legitimately slow and are not capped by this value.
|
|
|
|
|
|
const timeoutStr = process.env.IMMICH_TIMEOUT_MS;
|
|
|
|
|
|
const timeoutMs =
|
|
|
|
|
|
timeoutStr && !isNaN(parseInt(timeoutStr, 10)) ? parseInt(timeoutStr, 10) : 15000;
|
|
|
|
|
|
|
|
|
|
|
|
// Manual cache-buster for image and video URLs. Empty by default, so normal
|
|
|
|
|
|
// operation is untouched.
|
|
|
|
|
|
//
|
|
|
|
|
|
// Image responses are served `immutable` for a year, which means the browser
|
|
|
|
|
|
// does not revalidate at all — not even the ETag is consulted. That is the
|
|
|
|
|
|
// right default (these URLs are content-addressed by an encrypted asset ID),
|
|
|
|
|
|
// but it leaves no way to push a changed rendition: Immich regenerates
|
|
|
|
|
|
// thumbnails and previews when a job is re-run or a photo is rotated, and the
|
|
|
|
|
|
// asset ID — hence the token, hence the URL — stays the same. Bumping this
|
|
|
|
|
|
// value changes every image URL at once and is the only escape hatch short of
|
|
|
|
|
|
// rotating AUTH_SECRET.
|
|
|
|
|
|
const imageCacheVersion = (process.env.IMAGE_CACHE_VERSION || '').trim();
|
|
|
|
|
|
|
2026-03-08 21:12:52 +01:00
|
|
|
|
const rateLimitStr = process.env.RATE_LIMIT_RPM;
|
2026-03-12 11:30:45 +00:00
|
|
|
|
const rateLimit =
|
chore: clean up repo root and improve .gitignore
- Remove debug scripts: check_albums.js, check_albums.ts
- Remove git-filter-repo binary from root
- Untrack server.log (was accidentally committed)
- Delete .DS_Store files
- Rewrite .gitignore with clear sections: deps, next.js,
TypeScript, testing, logs, env, OS/editor, agent tools,
one-off scripts, deployment, user content
- Add .Jules/ and *.log glob to .gitignore
2026-03-13 10:33:07 +01:00
|
|
|
|
rateLimitStr && !isNaN(parseInt(rateLimitStr, 10)) ? parseInt(rateLimitStr, 10) : 1500;
|
2026-03-08 21:12:52 +01:00
|
|
|
|
|
2026-07-20 18:57:57 +02:00
|
|
|
|
// Number of reverse proxies in front of the app. Determines how far from the
|
|
|
|
|
|
// right of X-Forwarded-For the real client IP sits. nginx/Traefik/Caddy = 1;
|
|
|
|
|
|
// add 1 for each additional layer (e.g. Cloudflare in front of nginx = 2).
|
|
|
|
|
|
const hopsStr = process.env.TRUSTED_PROXY_HOPS;
|
|
|
|
|
|
let trustedProxyHops = hopsStr && !isNaN(parseInt(hopsStr, 10)) ? parseInt(hopsStr, 10) : 0;
|
|
|
|
|
|
|
|
|
|
|
|
// TRUSTED_PROXIES (removed) matched proxy IPs against the socket peer address,
|
|
|
|
|
|
// which self-hosted Next.js never exposes — so it silently disabled per-client
|
|
|
|
|
|
// rate limiting and lumped every visitor into one shared bucket. Migrate the
|
|
|
|
|
|
// common single-proxy case rather than leaving those deploys broken.
|
|
|
|
|
|
if (trustedProxyHops === 0 && process.env.TRUSTED_PROXIES) {
|
|
|
|
|
|
trustedProxyHops = 1;
|
|
|
|
|
|
console.warn(
|
|
|
|
|
|
'\n⚠️ TRUSTED_PROXIES is deprecated and has no effect. Assuming TRUSTED_PROXY_HOPS=1.\n' +
|
|
|
|
|
|
' Set TRUSTED_PROXY_HOPS explicitly (nginx/Traefik/Caddy = 1) and remove TRUSTED_PROXIES.\n',
|
|
|
|
|
|
);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
if (trustedProxyHops < 0) trustedProxyHops = 0;
|
|
|
|
|
|
|
2026-03-08 21:12:52 +01:00
|
|
|
|
return {
|
|
|
|
|
|
IMMICH_API_URL: apiUrl,
|
|
|
|
|
|
IMMICH_API_KEY: apiKey as string,
|
|
|
|
|
|
SITE_TITLE: process.env.SITE_TITLE || 'Gallery',
|
|
|
|
|
|
SITE_SUBTITLE: process.env.SITE_SUBTITLE || '',
|
|
|
|
|
|
CACHE_TTL: Math.max(0, cacheTtl),
|
Close the gaps left after v0.9.2's outage handling: route tests, stale cache, loading states (#397)
* docs: brainstorm resilience/adoption features and plan follow-up work
Brainstorm and deep-dive on where Immich Folio breaks under Immich
outages and what blocks self-hosted adoption, plus an implementation
plan for the remaining gaps after v0.9.2 (route-handler tests,
stale-while-error caching, loading skeletons).
* test: enable route-handler tests and cover admin guards and image content-type
vitest.config.ts only ran lib/__tests__/**, so route handlers were
structurally untestable. Two mechanisms live only at the route level and
cannot be extracted the way lib/imageSize.ts was: the admin auth guards
and the image proxy's Content-Type sanitisation (the stored-XSS fix from
c2fa8e7). Both were correct but unverified by any test.
Adds a table-driven test over all ten guarded /api/admin/* handlers,
including a filesystem walk that fails if a future route is added
without a row in the table. Adds a route-level test for
/api/image/[id] covering token validation, the six Content-Type
rewrite cases, ETag/304, and the 404 vs 503 (ImmichUnavailableError)
distinction.
Verified both catch real regressions by temporarily removing the admin
guards and the Content-Type rewrite and confirming the expected tests
fail, then restoring and confirming green again.
134 new tests (244 total, up from 210), tsc 0 errors, build succeeds.
* feat: serve stale cache entries while Immich is unavailable
v0.9.2 (#394) correctly split "gone" (404/410) from "unavailable"
(ImmichUnavailableError), and pages now render a 5xx instead of a false
404 during an outage — but that 5xx still replaced a gallery that had
loaded fine seconds earlier. lib/cache.ts deleted expired entries on
access, so nothing survived long enough to fall back to.
Cache entries now carry two deadlines: staleAt (the existing CACHE_TTL,
after which get() still reports a miss) and hardExpiresAt (the new
STALE_MAX_AGE, default 24h, after which the entry is dropped entirely).
Between the two, getStale() can still return the data.
getAlbums(), getAlbum() and getAssetInfo() catch ImmichUnavailableError
and fall back to getStale() before propagating. The two MISSING-sentinel
writes (definitive 404s) are deliberately left on the short cacheTtl
window — a corrected asset ID should take effect after the normal TTL,
not sit in the 24h stale window meant for outages.
STALE_MAX_AGE=0 disables the fallback. Verified the fallback only
engages on a real outage, not on a cache hit at the TTL boundary, and
that a definitive 404 still reports missing rather than falling back.
256 tests (up from 244), tsc 0 errors, build succeeds.
* feat: add loading skeletons for dynamic gallery routes
Every gallery route is force-dynamic and waits on a live Immich
round-trip — up to IMMICH_TIMEOUT_MS when the server is struggling —
with nothing rendered until it resolves. loading.tsx renders instantly
while the page component awaits its data.
Not browser-verified against live data: this worktree has no
.env.local and only .example content files, so the app boots into
setup mode (confirmed the dev server starts cleanly and the setup
screen renders, HTTP 200). tsc 0 errors, build succeeds, no new lint
errors.
* docs: document stale cache window, image token revocation, route-test convention
- README: STALE_MAX_AGE env var and a "Behaviour when Immich is
unreachable" section (stale-serves-until-STALE_MAX_AGE, 503 not 404,
outages never cached).
- docs/gallery-config.md: new "Image Token Revocation" subsection under
System & Security — un-publishing an album does not revoke already
issued image URLs; rotating AUTH_SECRET does.
- CLAUDE.md: updated cache description to cover the stale window and
MISSING sentinel, and a new code-convention note that route-level
logic (guards, header sanitisation) should get a route-level test
rather than being extracted into lib/ just to reach the test runner.
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
2026-08-07 11:07:34 +02:00
|
|
|
|
STALE_MAX_AGE: Math.max(0, staleMaxAge),
|
fix: Immich outage handling, image tiers, admin save safety (#394)
* fix: migrate deprecated middleware convention to proxy
Next.js 16 deprecated the `middleware` file convention in favour of
`proxy`. The build warned on every run, and the convention is slated for
removal in a future major — at which point Next.js would silently stop
invoking this file. That is not a cosmetic failure: proxy.ts owns the
Content-Security-Policy and the per-request nonce, so an uninvoked file
means every document response ships with no CSP at all, with nothing in
the build output to catch it.
Renames middleware.ts to proxy.ts and the exported middleware() to
proxy(), per the official migration. The config.matcher export is
unchanged, so route coverage is identical.
Adds lib/__tests__/proxy.test.ts, which pins the export name and
verifies the behaviour a silent regression would take out: a CSP is set,
the nonce in the policy matches the one forwarded to pages via x-nonce,
nonces differ per request, x-pathname still reaches the root layout, and
the matcher continues to cover /admin.
Verified: tsc 0 errors, 105 unit tests passing (98 before), production
build succeeds with 0 deprecation warnings.
* fix: stop reporting Immich outages as missing albums
lib/immich.ts request() returned null for every failure mode: a 5xx, a
rejected API key, a network error, and a non-JSON body all looked exactly
like "Immich answered and said this does not exist". Callers had no way
to tell the two apart, and app/[...path]/page.tsx calls notFound() on a
null album in five places.
The result was that while Immich was down or returning 5xx, every album
URL served a hard 404 — for albums that exist. Visitors saw "not found",
and crawlers were free to de-index real content over a transient outage.
request() now throws ImmichUnavailableError for anything that is not a
definitive answer about the resource, and reserves null for 404 and 410.
Pages let that propagate, so an outage surfaces as a 500 rather than a
404. Two callers must not propagate and are handled explicitly:
- ping() catches and returns false — it answers "is Immich reachable",
and both callers (/api/health, /api/admin/status) render that as a
status field.
- /api/exif and /api/map return 503 with Retry-After instead of 404 and
an empty marker list respectively. An empty map is otherwise
indistinguishable from a gallery with no GPS data.
Adds 10 tests: each failure mode maps to the right outcome, getAlbum and
getAlbums propagate rather than reporting empty, ping stays non-throwing,
and a failed request leaves no stale entry in the in-flight coalescing
map — which would otherwise replay the failure after Immich recovered.
Verified: tsc 0 errors, 115 unit tests passing (105 before), production
build exit 0, eslint 0 errors on all touched files.
* fix: bound how long Immich requests may hang
lib/immich.ts set no timeout on any of its three fetch call sites, so the
only ceiling was undici's built-in default, measured in minutes. Every
page is force-dynamic, which means an Immich that accepts connections but
never answers held each visitor's render open for that whole window —
and since nothing caches the failure, load kept arriving.
Adds IMMICH_TIMEOUT_MS (default 15000, floored at 1000 so a stray 0
cannot mean "no timeout"), threaded through env.ts and AppConfig like
CACHE_TTL.
The two kinds of call need different treatment:
- request() carries a whole-exchange AbortSignal.timeout(). JSON
payloads are small, so one budget can cover headers and body.
- streamAsset()/streamVideo() use an AbortController whose timer is
cleared in a finally the moment response headers arrive. A large
original photo or a long video is legitimately slow to transfer; a
whole-request timeout would truncate playback and break seeking, since
every range request would restart the clock.
A timed-out request surfaces as ImmichUnavailableError naming the budget,
so a misconfigured value is diagnosable from the log line alone.
Adds 5 tests, including the two that pin the distinction: a stream whose
headers have arrived is not aborted an hour later, while a stream that
never sends headers still is.
Verified: tsc 0 errors, 120 unit tests passing (115 before), production
build exit 0, no new eslint errors (the one error in env.ts is
pre-existing — confirmed against a clean tree).
* fix: stop caching Immich outages as permanently missing media
f3 split "gone" from "unavailable" in request(), but streamAsset() and
streamVideo() have their own fetch calls and were left returning null for
every failure. Both proxy routes turn null into a 404.
That 404 carried no Cache-Control, and a 404 is heuristically cacheable
per RFC 9111 — on URLs the success path serves as `immutable`. So a
transient outage, or a timeout introduced in 370e668, could pin a broken
image or an unplayable video in the visitor's browser cache well after
Immich recovered.
Both stream methods now throw ImmichUnavailableError for transport
failures, timeouts, and any error status other than 404/410, and reserve
null for genuinely absent media. A 200 with no body counts as
unavailable, not as missing. The routes translate that into 503 with
Retry-After, and both error responses — the 503 and the real 404 — now
send Cache-Control: no-store.
streamVideo keeps accepting 206, so range requests and seeking are
unaffected.
Adds 7 tests covering each failure mode for both methods plus the 206
path, and updates the F4 timeout test, which asserted the old null return.
Verified: tsc 0 errors, 127 unit tests passing (120 before), production
build exit 0, eslint 0 errors on touched files.
* perf: cache "does not exist" answers from Immich
getAlbum() and getAssetInfo() cached successes but discarded a definitive
"not found", so a missing resource was re-queried upstream on every call.
Measured before: 5 sequential lookups of a missing asset produced 5
upstream fetches. After: 1.
This is not hypothetical. app/page.tsx resolves every gallery.yaml hero
ID through getAssetInfo() on each render, and the homepage is
force-dynamic — so a hero photo deleted from Immich but still listed in
gallery.yaml cost one 404 round-trip per page view, indefinitely.
Absence is stored as a frozen MISSING sentinel because the cache cannot
distinguish a miss from a stored null; both read back as null.
Only definitive 404/410 answers are cached. An outage throws
ImmichUnavailableError and is deliberately never stored — caching one
would keep the gallery broken after Immich recovered. A test pins that:
a 503 followed by a healthy response must reach Immich the second time.
Stored under the normal cacheTtl rather than a separate shorter negative
TTL, because the reason for a shorter one no longer applies: outages are
not cached at all now, and both invalidation paths (the Immich webhook
and the admin save/reload) clear the entry immediately, so a corrected
asset ID takes effect at once rather than waiting out a TTL.
Also raises the test config's cacheTtl from 0, which expired entries in
the same millisecond they were written and would have made any caching
assertion flaky. Test isolation comes from cache.clear() in beforeEach.
Verified: tsc 0 errors, 131 unit tests passing (127 before), production
build exit 0, eslint 0 errors on touched files.
* fix: add themed error and not-found pages
The app had no app/error.tsx, app/not-found.tsx or global-error.tsx, so
every failure fell through to Next.js's unstyled default page.
That gap became visible with 253ade0, which made an Immich outage produce
a 500 instead of a false 404 — correct, but what a visitor saw was a bare
framework error screen with none of the site's theming.
The two pages also carry the distinction the error split was for, and
they carry it structurally rather than by inspecting the error: notFound()
renders not-found.tsx, a throw renders error.tsx. That matters because
Next.js strips error messages in production and replaces them with a
digest — this component genuinely cannot tell an Immich outage from any
other server error, so it does not claim to. It frames the failure as
temporary and offers reset(); not-found.tsx deliberately does not, since
a real 404 is not transient.
The digest is shown so a user report can be traced to a log line. The raw
error message is never rendered — it can carry an upstream URL, a header
or a config value.
Both pages reuse .empty-state, which existed in globals.css but was
unreferenced anywhere in the app; extended with title, actions and button
elements matching the PasswordGate button, the only other action button
in the codebase.
Tested by server-rendering both components (react-dom/server, no new
dependencies) — 8 tests covering the styling hook, the retry and home
affordances, digest handling, and that a message containing an API key or
host address does not reach the markup. Note vitest.config.ts only
includes lib/__tests__/**/*.test.ts, so the test uses createElement
rather than JSX; a .tsx file there would never run.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the app boots into setup mode and neither
Playwright nor a dev server can exercise these routes here.
Verified: tsc 0 errors, 139 unit tests passing (131 before), production
build exit 0 with /_not-found registered, eslint 0 errors.
* perf: drop the unused originalUrl from every photo item
PhotoItem declared originalUrl and toPhotoItems computed it for every
asset, but a repo-wide grep found exactly two references: the field
declaration and the assignment. Nothing ever read it — the grid uses
thumbUrl, the lightbox uses previewUrl.
The cost was not the encryption. PhotoItem is a server-to-client prop, so
the dead field was serialized into the RSC payload and shipped to the
browser for every photo.
Measured: 137 bytes per asset once serialized, so 6.7 KB for a 50-photo
album and 66.9 KB for a 500-photo one — transferred and parsed on every
page view, never read.
Removed rather than wired up to a "view original" affordance. Exposing
full-size originals is a bandwidth and privacy decision for the project
to make deliberately, not something to infer from a dangling field.
No unit test: this is a deletion whose correctness is exactly "nothing
referenced it", which tsc and a repo-wide grep already establish. A test
asserting the absence of a field would restate the type checker.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint unchanged at 4 pre-existing errors on the touched files (unused
imports, confirmed identical against a stashed clean tree).
* chore: remove four orphaned imports from the catch-all route
Link, Image, PhotoGrid and BackLink were imported by app/[...path]/page.tsx
and never used — 4 of the 6 no-unused-vars errors in the repo baseline.
Verified they are leftovers, not a dropped render path: PhotoGrid and
BackLink are both rendered by AlbumDetailView, which imports them itself,
and BackLink is also used by the impressum and map pages. Rendering was
extracted out of page.tsx into AlbumDetailView/SubpageGridView at some
point and the imports stayed behind. `type PhotoItem` is still needed by
toPhotoItems, so that import narrows to a type-only import rather than
disappearing.
No bundle win, measured: .next/static is 1084 KB across 20 JS chunks both
before and after, byte-identical. Turbopack already tree-shook them. The
backlog note that predicted unused client-component imports would hold
them in the client graph was wrong, and has been corrected there.
The value is readability and lint: repo errors drop from 54 to 50, and
page.tsx no longer reads as though it renders the grid itself.
chore: rather than fix:/perf: — nothing behaves differently and the
measurement showed no performance change, so either of those labels would
misdescribe it. CLAUDE.md lists chore: as a valid prefix.
Verified: tsc 0 errors, 139 unit tests passing, production build exit 0,
eslint 0 errors on the touched file.
* fix: make ?w= able to narrow the image tier instead of being ignored
The image proxy resolved its Immich rendition with `?size=` winning
whenever present, falling back to `?w=` otherwise. lib/urls.ts always
writes `?size=`, so for every URL this app generates the width branch was
unreachable and widthToSize() was dead code — while CLAUDE.md documented
the width mapping as the actual behaviour.
`?size=` is now a ceiling rather than an override: when both are present
the smaller tier wins, so a width can narrow the request but never widen
it. That asymmetry is the point. Letting width win outright — the obvious
reading of the docs — would have been a real regression: next/image emits
widths up to 3840 and widthToSize(1920) is 'original', so every large
display would have started downloading full-size originals in place of
previews. A test pins that specific case.
No measured change today. The only URLs carrying both parameters are
?size=preview with next/image widths, whose smallest vw-based deviceSize
is 640 — already preview either way. This is a correctness fix that makes
the documented contract true and testable, not a performance win.
Moved the logic to lib/imageSize.ts so it can be tested at all:
vitest.config.ts only collects lib/__tests__, so nothing inside a route
file is reachable from the suite. 9 tests cover the tier boundaries, both
single-parameter paths, malformed input, and the ceiling behaviour.
CLAUDE.md updated to describe the ceiling rule rather than a plain
width mapping.
Verified: tsc 0 errors, 148 unit tests passing (139 before), production
build exit 0, eslint 0 errors on touched files.
* fix: validate the per-album heroImage UUID from gallery.yaml
Every asset ID in gallery.yaml is routed through validateUuid before it
can reach encodeAssetId, which encrypts whatever it is handed. One was
not: for a per-album entry the album ID (the map key) was validated while
the heroImage value was stored raw.
A typo there produced a perfectly well-formed image URL that 404s forever,
with nothing in the log naming the cause — the failure surfaced as a
broken hero image on an album page, far from the line that caused it.
This is narrower than the finding that prompted it, which proposed
guarding encodeAssetId itself. Validation belongs at config load, where it
runs once and can name the file and the album; a guard in encodeAssetId
would run on every render and duplicate validateUuid at the wrong layer.
The heroImage path was the only unvalidated route into it.
validateUuid warns and substitutes the nil UUID rather than throwing, so a
bad ID still degrades to a missing image instead of taking the page down —
and since f511e62 that 404 is negative-cached, so it costs one upstream
request rather than one per render.
Adds 4 tests. The two that pin the wiring were mutation-checked: both fail
against the unfixed code and pass with it.
Verified: tsc 0 errors, 152 unit tests passing (148 before), production
build exit 0, eslint 0 errors on touched files.
* fix: keep /admin reachable when gallery.yaml cannot be derived
getConfig() throws on three shapes the admin page builder is able to
write: a gallery with neither albums nor subpages, a subpage with no name,
and a subpage with neither albums nor sections.
RootLayout called it unguarded, and a throw in a *layout* is not caught by
app/error.tsx — that needs a global-error boundary, which this app does
not have. /admin renders inside that same layout, so one bad save took
down the site together with the only tool that could undo it. Recovery
meant editing content/gallery.yaml on the host, or exec'ing into the
container. This is issue #326's failure mode reached through an invalid
config instead of a missing one.
Adds getConfigOrNull(), which returns null rather than throwing and logs
the underlying error plus the path to content/.backups/. The layout treats
null exactly like needsSetup: setup screen for visitors, children rendered
for admin paths so the panel still loads. generateMetadata() is guarded
too — it also calls getConfig(), and it also runs before any error
boundary.
This is the half of the finding that makes every broken config
recoverable, including ones this app never wrote: a hand-edited file, a
restored backup, a config from an older version. The other half —
validating in the admin PUT so the bad state is never written — is filed
separately; it prevents new breakage but cannot repair existing breakage,
so it was not the one to do first.
Adds 10 tests. Three pin the premise by asserting getConfig() really does
throw on each shape, so the fallback cannot quietly become dead weight.
One asserts the needsSetup path still returns a config rather than null,
since missing and invalid are different states that must not collapse.
Not verified in a browser: this worktree has no .env.local and only
.example content files, so the layout/error-boundary interaction is
reasoned from Next.js semantics and covered at the getConfig level only.
Verified: tsc 0 errors, 162 unit tests passing (152 before), production
build exit 0, eslint 0 errors on touched files.
* fix: reject an admin save the site would not be able to load
The admin PUT validated only the shape of the hero field, then wrote. None
of getConfig()'s three throw conditions was checked, so the page builder
could persist a gallery.yaml that cannot be loaded — and the user got
"Saved successfully" for it. Since 95877af that no longer locks anyone
out, but the public site still drops to the setup screen until someone
notices.
Extracts the gallery half of getConfig() into deriveGallery(), which the
route now runs against the proposed gallery before writing. Two
alternatives were rejected on purpose:
- Re-checking the three conditions in the route would put a second copy of
the rules next to the first, and the two would drift.
- Write-then-verify-then-roll-back leaves a window in which other requests
read the broken config.
The extraction is a pure move: no logic changed inside it, and all 162
pre-existing tests pass untouched, which is what makes that claim
checkable rather than asserted.
The 400 carries the derivation's own message, so the admin UI names the
offending subpage instead of saying "invalid gallery" — a builder with a
dozen subpages needs to know which one. A test pins that the message
identifies it.
Adds 8 tests: each rejected shape, the named-subpage message, and the
valid structures (standalone albums, subpage membership, sections,
per-album overrides and hero images) so the extraction cannot silently
lose behaviour.
Verified: tsc 0 errors, 170 unit tests passing (162 before), production
build exit 0, eslint 0 errors on touched files.
* fix: hash subpage passwords off the main thread
verifyScrypt used crypto.scryptSync, measured at 22.8ms per call — 22.8ms
during which the Node process serves nothing else, because the sync
variant runs on the main thread rather than the libuv threadpool.
The /api/auth rate limit does not contain this. It allows 10 attempts per
minute, but the bucket key is the client IP, and with the default
TRUSTED_PROXY_HOPS=0 that IP is read from x-real-ip / x-forwarded-for —
both attacker-controlled. lib/rate-limit.ts already documents that those
headers are spoofable at hops=0; what followed from it was not written
down anywhere: one attacker rotating a header gets unlimited buckets, and
each request buys 22.8ms of dead process. Roughly 44 requests per second
saturates it.
Switches verifyScrypt and generateScryptHash to the async crypto.scrypt,
which runs on the threadpool. This does not stop the spoofing — that is
inherent to hops=0 and already documented — it removes the amplification
that made spoofing worth doing.
authenticate() becomes async. It had exactly one caller
(app/api/auth/route.ts). isAuthenticated() is deliberately left
synchronous: it only computes an HMAC, and it is called from server
components.
Adds 4 tests, two of which pin the fix and were mutation-checked against
the old implementation — both fail with scryptSync and pass with scrypt.
One asserts scryptSync is never called; the other asserts a 2ms interval
still fires during a hash, which is the property that actually matters. A
slower machine lengthens the hash and yields more ticks, so that test
fails on a blocked loop rather than on a busy one.
Also adds a scrypt-hashed fixture subpage, so the suite exercises
verifyScrypt rather than only the deprecated plaintext fallback.
Verified: tsc 0 errors, 174 unit tests passing (170 before), production
build exit 0, eslint 0 errors on touched files.
* fix: warn when TRUSTED_PROXY_HOPS cannot identify a client
getClientIp falls back to a shared 'unknown' bucket whenever the
X-Forwarded-For chain is shorter than the configured hop count, or the
header is absent with hops > 1.
That fallback is correct — the alternative is trusting a spoofable entry —
but it was silent, and silence is the problem. Every unidentified request
shares one bucket per endpoint, and this is an image proxy: a single
50-photo grid issues ~50 /api/image requests, so the default
RATE_LIMIT_RPM=1500 is collectively spent by roughly 30 page loads a
minute. Someone who sets TRUSTED_PROXY_HOPS=2 for Cloudflare-plus-nginx
and then reaches nginx directly gets a site that 429s everyone, with
nothing in the log pointing at the cause.
Warns once per process, naming the configured hop count and what was
actually seen — the chain length, or which header was missing. Once, not
per request: enough to diagnose without flooding the log.
The fallback itself is unchanged. Trusting a shorter chain would hand
every client its own bucket key and defeat the limiter entirely.
No warning at hops=0, where the shared bucket is the documented
consequence of not declaring a proxy rather than a misconfiguration.
Adds 6 tests covering each fall-through path, the once-per-process
behaviour, and both silence cases. The once-per-process flag needs a test
seam to re-arm, hence the __resetProxyWarningForTests export.
Verified: tsc 0 errors, 180 unit tests passing (174 before), production
build exit 0, eslint 0 errors on touched files.
* fix: send Retry-After on every rate-limited response
/api/image, /api/video and /api/exif each computed the retry delay,
interpolated it into a console.warn, and then returned a 429 without it —
the number was calculated and thrown away. /api/og did not compute it at
all. Only /api/auth and /api/map sent the header.
This matters more for an image proxy than for a typical API: one gallery
page issues around fifty /api/image requests, so a burst that trips the
limit produces fifty rejections carrying no back-off signal. The browser
and next/image then retry on their own schedule — a retry storm against a
server that just reported it was overloaded.
Adds Retry-After and Cache-Control: no-store to all four. no-store because
the success path serves these same URLs as `immutable`.
The delay now comes from a shared retryAfterSeconds() helper rather than
an inline Math.ceil in each route, which also fixes an edge every copy
shared: a window on the point of expiring yields 0, and Retry-After: 0
tells the client to retry immediately, which is the opposite of what a 429
means. A resetAt already in the past yields a negative number, which is
not valid HTTP. The helper floors at 1.
The three routes that already sent the header (/api/auth, /api/map,
/api/admin/auth) and /api/health now use the helper too — leaving four
copies of the buggy arithmetic behind while centralising the rest would
have been worse than not centralising at all.
Adds 2 tests, one of them specifically for the zero-and-negative edge.
Verified: tsc 0 errors, 182 unit tests passing (180 before), production
build exit 0, eslint 0 errors on all ten touched files.
* fix: add an escape hatch for stale browser image caches
Image and video responses are served
`Cache-Control: public, max-age=31536000, immutable`, so a browser does
not revalidate them for a year — the ETag is not even consulted. When
Immich regenerates thumbnails after a job re-run, or a photo is rotated,
the asset ID does not change, so neither does the deterministic token or
the URL. Visitors keep the old rendition, and the webhook that clears the
server-side caches cannot reach a browser. The only existing remedy was
rotating AUTH_SECRET, which invalidates every token on the site.
Corrects the framing this started from. The finding described it as an
ETag problem, but `immutable` means the ETag is never reached in the case
that matters. The URL is the only lever a browser cache responds to, so
the buster belongs there. It is also mixed into the ETag, since a matching
ETag across two different URLs would be wrong regardless.
Adds IMAGE_CACHE_VERSION, empty by default. Unset, the generated URLs are
byte-identical to before — a test asserts exactly that, because a
cache-busting feature that changes URLs when nobody asked for it would
silently discard every visitor's cache on deploy.
The value is URL-encoded at parse time; an unescaped one could otherwise
inject query parameters into every image URL on the site.
Deliberately not done: shortening max-age or dropping `immutable`. Both
trade a large, real bandwidth win for a rare correctness case, and the
cost is not measurable from this worktree. The aggressive caching stays
the default; this only adds a way out.
Adds 6 tests: default-unchanged, whitespace treated as unset, both URL
shapes, escaping, and that a bump changes the query while leaving the
token intact.
Verified: tsc 0 errors, 188 unit tests passing (182 before), production
build exit 0, eslint unchanged (the one error in env.ts is pre-existing,
confirmed against a stashed tree).
* security: validate the backup filename before restoring from it
restoreBackup() joined its argument straight into a path and copied the
result over a content file. path.join resolves `..`, so
restoreBackup('../../../../etc/passwd') would have written any readable
file over content/settings.yaml — whose contents the admin GET endpoints
hand straight back. The destination was chosen by
backupFilename.includes('gallery.yaml'), which is substring matching on
the same attacker-supplied string, so the caller picked that too.
Not exploitable today: the function has zero callers. That is precisely
why it was worth fixing rather than leaving — it is an exported function
carrying a traversal bug, waiting for whoever wires up the restore button
the admin panel does not yet have.
Hardened rather than deleted. The status endpoint already reports backup
counts, so restore is plainly intended; removing a half-built feature is
the maintainer's call, not a side effect of a security fix.
The name must now match the pattern this service itself produces, and the
destination is derived from the matched group rather than a substring
search. The redundant basename check is kept on purpose: it preserves the
guarantee if that pattern is ever loosened.
Adds 12 tests. Nine of them fail against the unguarded version —
mutation-checked by removing the guard and confirming, rather than
assuming they would.
Verified: tsc 0 errors, 200 unit tests passing (188 before), production
build exit 0, eslint 0 errors on touched files.
* fix: give each admin save its own temp file
writeYamlFile wrote to a constant `${filePath}.tmp` before renaming. The
rename is what makes the write atomic; the temp filename was not. Two
saves of the same file in flight at once shared that path, so the second
writeFile could interleave with the first's rename — publishing one save's
bytes under the other's, or a truncated mix of both. A double-clicked Save
in the page builder is enough to reach it.
The temp name now carries the pid and a per-process counter, so concurrent
writers cannot collide either within a process or across workers.
Also wraps the write and rename so a failure unlinks the temp file instead
of leaving it in content/. The cleanup swallows its own errors: failing to
tidy up must not replace the real error with a misleading one.
Adds 7 tests, and both halves were mutation-checked independently —
reverting to the fixed temp path fails two of them, removing the cleanup
fails two others. Two of the tests exist to pin properties that are easy
to regress silently: that the rename targets the temp file just written,
and that the happy path issues no unlink.
Verified: tsc 0 errors, 207 unit tests passing (200 before), production
build exit 0, eslint 0 errors on touched files.
* fix: await the stream-timeout rejection before advancing timers
`npm run test:unit` exited 1 even though all 207 tests passed. The
timeout test started `streamAsset`, advanced the fake timers past the
15s budget, and only then attached the rejection expectation. The abort
fires *during* `advanceTimersByTimeAsync`, so for one microtask turn the
promise was rejected with no handler attached — Node reports that as an
unhandled rejection, and vitest fails the whole run on it regardless of
test outcomes. CI would have blocked on a green suite.
Attaching the assertion before advancing the timers fixes the ordering
without weakening it: the same rejection and the same `captured.aborted`
check still run, just awaited after the advance rather than created
after it.
Verified: vitest exit 0, 207/207 passing, tsc 0 errors, prettier clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-05 12:04:01 +02:00
|
|
|
|
// Kept URL-safe: this value is appended to every image and video URL.
|
|
|
|
|
|
IMAGE_CACHE_VERSION: encodeURIComponent(imageCacheVersion),
|
|
|
|
|
|
// A sub-second budget would abort healthy requests; 0 must not mean "no timeout".
|
|
|
|
|
|
IMMICH_TIMEOUT_MS: Math.max(1000, timeoutMs),
|
2026-03-08 21:12:52 +01:00
|
|
|
|
RATE_LIMIT_RPM: Math.max(1, rateLimit),
|
|
|
|
|
|
AUTH_SECRET: process.env.AUTH_SECRET,
|
2026-07-20 18:57:57 +02:00
|
|
|
|
TRUSTED_PROXY_HOPS: trustedProxyHops,
|
feat: add visual admin panel with page builder and settings editor
- Add /admin route with password-protected visual page builder
- Page Builder: manage hero images, standalone albums, subpages, and sections
- Album Picker: browse all shared Immich albums with search and filtering
- Settings Editor: theme, grid, footer, legal, SEO configuration
- YAML service with atomic writes and automatic backups (content/.backups/)
- HMAC-signed admin sessions (24h expiry, HttpOnly cookies)
- API endpoints: /api/admin/{auth,gallery,settings,albums,reload}
- Enable with ADMIN_PASSWORD env var; disabled by default
- invalidateConfigCache() export for hot-reload after saves
2026-05-29 17:04:37 +02:00
|
|
|
|
WEBHOOK_SECRET: process.env.WEBHOOK_SECRET || undefined,
|
|
|
|
|
|
ADMIN_PASSWORD: process.env.ADMIN_PASSWORD || undefined,
|
release: v0.11.0 — journal fixes, admin URLs, design system cleanup (#432)
* fix(screenshots): sharp-Typen fuer 0.34 und 0.35 gleichzeitig (#410)
sharp 0.35 ist von `export =` auf ESM umgestellt: der Modul-Namespace
ist nicht mehr aufrufbar, die Factory sitzt unter .default. Der
Type-Check brach damit in scripts/screenshots.ts weg, sobald der
Next-Bump (#402) sharp von 0.34.5 auf 0.35.3 mitzieht -- sharp ist hier
keine deklarierte Dependency, sondern kommt ueber Next mit.
Die Form wird jetzt aus dem installierten Paket abgeleitet statt
festgeschrieben, zur Laufzeit wie im Typ. Geprueft mit beiden
Versionen: tsc laeuft mit 0.34.5 und mit 0.35.3 fehlerfrei durch.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore(deps): bump the nextjs group across 1 directory with 2 updates (#402)
Bumps the nextjs group with 2 updates in the / directory: [next](https://github.com/vercel/next.js) and [eslint-config-next](https://github.com/vercel/next.js/tree/HEAD/packages/eslint-config-next).
Updates `next` from 16.2.12 to 16.3.0
- [Release notes](https://github.com/vercel/next.js/releases)
- [Commits](https://github.com/vercel/next.js/compare/v16.2.12...v16.3.0)
Updates `eslint-config-next` from 16.2.12 to 16.3.0
- [Release notes](https://github.com/vercel/next.js/releases)
- [Commits](https://github.com/vercel/next.js/commits/v16.3.0/packages/eslint-config-next)
---
updated-dependencies:
- dependency-name: eslint-config-next
dependency-version: 16.3.0
dependency-type: direct:development
update-type: version-update:semver-minor
dependency-group: nextjs
- dependency-name: next
dependency-version: 16.3.0
dependency-type: direct:production
update-type: version-update:semver-minor
dependency-group: nextjs
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* chore(deps-dev): bump prettier from 3.8.3 to 3.9.6 (#405)
Bumps [prettier](https://github.com/prettier/prettier) from 3.8.3 to 3.9.6.
- [Release notes](https://github.com/prettier/prettier/releases)
- [Changelog](https://github.com/prettier/prettier/blob/main/CHANGELOG.md)
- [Commits](https://github.com/prettier/prettier/compare/3.8.3...3.9.6)
---
updated-dependencies:
- dependency-name: prettier
dependency-version: 3.9.6
dependency-type: direct:development
update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* docs(changelog): Unreleased-Sektion fuer die Bumps nach v0.10.0 (#413)
Die Dependency-PRs sind nach dem Release auf dev gelandet, teils auch
auf main. Ohne Eintrag waeren sie beim naechsten Release unsichtbar --
inklusive Next 16.3.0, das CVE-2025-13465 mitbringt.
Ausserdem im 0.10.0-Abschnitt nachgetragen, dass der Trivy-Scan bis
#409 nie gelaufen ist. Der Fix steckt im Tag v0.10.0, fehlte aber in der
Aufzaehlung -- und "Images werden bei jedem Publish gescannt" stand im
Changelog, waehrend der Job im Action-Setup abbrach.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(gallery): per-album sort order, independent of Immich (#408) (#414)
Album order was dictated entirely by Immich: getAlbum() sorted by the
album's own asc/desc flag, so a Folio album always mirrored the archive.
A curated series has a narrative order that has nothing to do with
capture time, and changing the Immich album's sort to fit the website
changes it for the archive too.
Adds an optional per-album `sort` in gallery.yaml, editable in /admin,
defaulting to today's behaviour so every existing config is unchanged:
immich (default) | newest | oldest | filename | manual
`manual` is a pinned *prefix*: `assetOrder` lists only the photos placed
by hand, everything else follows in the album's Immich order. A full
permutation of a 2,000-photo album would be ~80 KB of UUIDs in a file
that is rewritten and copied to content/.backups/ on every admin save.
It also makes drift fall out for free — photos removed in Immich never
match, new ones land in the tail rather than being hidden.
Sorting is applied after the cache read, not before the cache write, so
the LRU keeps one canonical Immich-ordered copy per album. That avoids
putting the mode in the cache key, lets a config change take effect
without depending on invalidation, and keeps the stale-serving path
correct. cache.get() returns the stored object by reference, so
withSort() returns a shallow clone with a new array — sorting in place
would reorder the shared entry, and under request coalescing every
concurrent caller would be sorting the same array.
Also fixes a prerequisite bug: /api/admin/albums/[albumId]/assets read
album.assets from GET /albums/:id, which Immich 3.x no longer populates
(lib/immich.ts:507). The hero picker's "This Album" tab returned nothing
and threw outright when the key was absent. It now goes through the
client's paginated metadata search, with ?types=all to opt videos in for
the reorder editor.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): drop decorative emoji from the sort select labels (#415)
An <option> cannot contain markup, so the SVG set in Icons is not
reachable there — but a decorative emoji is not a substitute for it.
Plain text instead; the album card badge already uses a real icon.
The remaining emoji in the admin panel (asset picker tabs, section
headings, layout picker options) predate this branch and are left for a
separate cleanup.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): replace emoji icons with the shared SVG icon set (#416)
The admin panel mixed two icon conventions: inline SVGs from
`Icons.tsx` in most components, and decorative emoji in the asset
picker and a few PageBuilder labels. Emoji render inconsistently
across platforms, ignore the accent colour, and cannot inherit the
`.svg-icon` sizing rules.
- Icons.tsx: add `IconStar` (filled — it marks a state, not an
action) and `IconImage`.
- AssetPicker: the three tab labels and the favourite tile badge now
use IconFolder / IconStar / IconImage.
- PageBuilder: the "Standalone Albums" and "Subpages" headings use
the icon set instead of an emoji.
- PageBuilder: the layout picker drops its emoji for plain text. An
<option> cannot contain markup, so an SVG is impossible there and a
decorative glyph is not a substitute.
- admin.css: align icon and label in the picker tabs, and swap the
favourite badge's text-shadow for drop-shadow, which is what
applies to an inline SVG.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* refactor(admin): fold PageBuilder's local icons into the shared Icons module (#418)
* refactor(admin): fold PageBuilder's local icons into the shared module
PageBuilder.tsx carried its own `Icons` object duplicating most of the
glyphs already exported by Icons.tsx, with hardcoded sizes and slightly
divergent paths. Migrate all 32 call sites to the shared module and
delete the local object.
Five glyphs had no counterpart and are added to Icons.tsx as 1:1 ports
of the local paths: IconX, IconPlus, IconHome, IconCopy, IconImage.
The local ExternalLink and Check are dropped rather than ported — both
had zero call sites.
Two glyphs change shape slightly, which is the point of consolidating:
the shared IconCamera includes the viewfinder bump the local copy
omitted, and drag handles now use IconGripVertical.
IconGripVertical uses the module's square 24x24 viewBox, so `.svg-drag`
goes from 12x18 to 18x18 — the old non-square box would letterbox the
glyph down to 12x12 and shrink the handle.
* fix(admin): resolve icon-module integration after #416
Merging dev brought in #416, which added a second IconImage to Icons.tsx
and two <Icons.Camera>/<Icons.Folder> section headers that reference the
local Icons object this PR removes. Git merged both cleanly but the result
did not type-check.
- drop the duplicate IconImage (both copies were byte-identical)
- point the two section headers at the imported IconCamera/IconFolder
- refresh the SORT_LABELS comment: the icon set is imported now, not
declared below
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* fix(config): allow an empty gallery, guard against a double /api suffix (#421)
An empty gallery is now a valid, rendered state instead of a throw: deriveGallery() no longer requires at least one album or subpage, so a deployment with no albums yet renders a bare homepage the owner can fill in via /admin.
getConfig() also stops appending /api to an Immich URL that already ends in it, which previously produced .../api/api and broke every upstream request.
Includes a restore of the per-album sort test cases, which the branch had lost by being cut before #408/#414.
* feat(admin): favicon upload, served from the content volume (#422)
Adds a bundled default favicon, an upload widget in admin Settings > General, and a route that serves the icon from the writable content/ volume with a fallback to the bundled default. Raster uploads (PNG/ICO/JPEG) are wrapped in an SVG container; the response carries its own CSP so an uploaded SVG cannot execute script on the origin.
The GET route revalidates with an ETag instead of caching immutably, so a newly uploaded icon shows up immediately. Uploads are written via temp file + rename, and a Reset button restores the bundled default.
* ci: flag unformatted files in a PR, without blocking it (#424)
Prettier is configured as an error rule in eslint.config.mjs but has never run in CI, so part of the tree predates it. A final CI step now diffs against the PR base, runs prettier --list-different over the changed files, and emits a warning annotation per offender plus a job summary.
Advisory by design: it never fails the job. Blocking would turn a PR red for merely touching a file that was already unformatted, and normalising the tree in one commit would wreck git blame and conflict with the PRs in flight.
Requires fetch-depth: 0, since diffing against the base needs its history.
* feat(install): first-run wizard gated by a one-time setup token (#419)
Adds /install, a four-step first-run wizard: connect to Immich, optionally pick albums, name the site, finish. It verifies the credentials against the Immich server before writing anything, then writes gallery.yaml, settings.yaml and a content/install.json carrying the Immich credentials and a generated site secret.
install.json exists because a running process cannot pick up new env vars — the wizard has to write configuration that takes effect without a restart. getInstallCredentials() merges it under the real environment, so env vars always win and rotation stays a matter of setting one.
Both the page and the API routes are gated by a one-time setup token printed to the server log, so a deployment that is reachable before gallery.yaml exists cannot be configured by whoever finds it first. The token is stored in content/.setup-token (0600) rather than a module variable: Next.js bundles the page and each route handler separately, so a module-level token left the page and the API disagreeing inside one process and the wizard stuck on step 1. It is removed once the install completes.
The admin password is stored as an scrypt hash using the scheme lib/auth.ts already applies to album passwords; a plaintext ADMIN_PASSWORD env var keeps working.
* docs: document the setup wizard, fix the read-only content mount (#425)
Quick Start now leads with the /install wizard, with the hand-written YAML route kept alongside it, and a First-Run Setup section covers the setup token, the files the wizard writes, and env-var precedence.
Two corrections: both Docker examples mounted content/ as :ro, which silently breaks the wizard, the admin panel and the backup rotation alike — docs/admin-panel.md already said so and the README contradicted it. And nothing documented that content/install.json holds the Immich API key, so a backup of content/ is a backup of credentials.
Also adds a Contributors section crediting the people whose work is in the project.
* feat(admin): floating save bar pinned to the bottom of the viewport (#426)
The save bar was the first, static element inside .admin-main. Since the
document itself scrolls, it left the viewport as soon as you scrolled down
— editing a subpage section or a settings block far below gave no visible
dirty hint and no reachable save button.
Replace it with a bottom-pinned floating bar shared by both editors. It
only mounts while there is something to act on (unsaved changes, an
in-flight save, or a result message), so it never covers content while
idle. The success message clears after 5s, which hides the bar again.
- new SaveBar component, replacing the duplicated markup in PageBuilder
and SettingsEditor
- adopt the previously unused .floating-save-bar CSS: slide-in keyframes
(carrying translateX(-50%)), dirty border, wrap/gap, mobile variant and
a prefers-reduced-motion opt-out; z-index 900 keeps it below modals
- .admin-main gains bottom padding so the bar never covers page content
- drop the now-unreachable idle hints: the ⌘S hint becomes a button title,
the Live Sync note becomes a static line above the settings layout
- header "Site" link now bypasses the cache (/?fresh=1), so the preview
stays reachable when the bar is hidden
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(admin): themed listbox for the album sort control (#427)
The per-album "Photo order" control was a native <select>. Because an
<option> cannot contain markup, its choices could not use the admin
panel's currentColor SVG set — the reason #415 had to fall back to plain
text after dropping the emoji labels.
Adds a Listbox component implementing the ARIA combobox/listbox pattern,
so options can render a glyph beside their label:
- DOM focus stays on the trigger button; the active option is announced
through aria-activedescendant, keeping focus management to one element
- Keyboard: arrows (clamped at the ends, like a native select),
Home/End, Enter/Space to commit, Escape to cancel, Tab commits and
moves on, and prefix type-ahead with a repeat-key cycle
- Outside dismissal on pointerdown, so a press starting outside closes
the popup before the click retargets
The sort options now carry an icon each: calendar, arrow-down, arrow-up,
an A→Z marker, and the drag glyph for manual — matching the badge the
album card already shows. SORT_LABELS stays plain text because the badge
feeds it to a title attribute.
Closes #417
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore(git): ignore generated design handoff folders
* fix(admin): stop the album sort dropdown clipping, and simplify the panel (#428)
* fix(admin): render listbox popup in a portal so it escapes modal clipping
The sort control's popup was position:absolute inside three clipping ancestors
(.album-drawer overflow:hidden > .album-drawer-body overflow-y:auto >
.modal-right-column overflow-y:auto), so it was cut off at the bottom of the
Edit Album Details modal. z-index could not help: clipping happens before
stacking. It also always hung below the trigger, with no flip.
Portal the popup to document.body and position it from the trigger's viewport
rect, flipping above when there is more room there and capping max-height to the
space available so it scrolls internally rather than overflowing. It is
dismissed on scroll and resize, since the anchor is only measured on open.
The outside-pointerdown handler now also treats the popup as inside; without
that the portalled list counts as outside, and a click on an option would close
it before the selection commits.
* refactor(admin): single-column album details panel
The left column spent 1.2fr on a cover thumbnail and three stat rows, two of
which (Status, Custom Hero) only restated the form fields opposite them, while
the form itself was squeezed into 1.8fr behind a second nested scroller.
Stack the drawer in one column instead: cover as a banner, then the form at full
width, with the UUID moved into the footer. The body is now the only scroll
container, which also removes one of the ancestors that was clipping the sort
popup.
---------
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
* fix(admin): give the portalled listbox its theme back, lock the page behind modals (#429)
Two defects in the album editor.
The sort listbox popup painted transparent. Every --admin-* variable is
scoped to .admin-layout, and since the popup moved into document.body it
sits outside that subtree, so background/border/text all resolved to
nothing. It now carries its own copy of the palette in both colour
schemes, plus the font it no longer inherits.
Scrolling inside a modal also moved the page behind it. No admin modal
ever locked body scroll: the overlay covers the page but does not absorb
wheel events, so a scroll starting on the backdrop — or continuing past
the end of the modal's own scroller — reaches the document.
overscroll-behavior cannot cover that case on its own, because it only
applies to scroll containers and the backdrop is not one.
useScrollLock freezes body while a modal is open and compensates for the
width the scrollbar gives up, so nothing shifts underneath. Each lock
restores the inline value it found, which makes nesting work without a
counter — a picker opened from the drawer restores the drawer's hidden
state instead of clearing it. Applied to both drawers and all four
picker modals, which share the defect. overscroll-behavior: contain on
the scrollers stops the chaining they can handle themselves.
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
* fix(security): allow unsafe-eval in CSP script-src during dev mode
* feat(journal): add Journal Studio with split-screen editor, live preview and dedicated routes
* fix(build): separate client parser from server journal-service and fix build type errors
* fix(journal): Fotos rendern, Layout und Lightbox-Steuerung korrigieren
Fünf unabhängige Fehler in der öffentlichen Journal- und Essay-Ansicht:
* Foto-Blöcke wurden nie angezeigt. EssayView löst ein Foto über
PhotoItem.id auf, und das sind verschlüsselte Tokens — die Blöcke aus
dem Markdown tragen aber rohe Asset-UUIDs. Jede Zuordnung lief ins
Leere, und nicht auflösbare Blöcke werden still verworfen. Die IDs
werden jetzt serverseitig übersetzt; damit verschwinden nebenbei die
rohen UUIDs aus dem Client-Payload, die dort laut Projektregel nie
auftauchen dürfen. Nicht auflösbare Referenzen werden geloggt.
* Überschriften standen weiter links als der Fließtext. Jeder Block
hatte einzeln `max-width: 68ch`, und `ch` ist relativ zur Schrift des
Elements — die Serif-Überschrift ergab eine breitere Spalte. Ersetzt
durch eine gemeinsame, font-unabhängige `--essay-measure`.
* Fullbleed-Fotos bluteten nur nach links aus: `.essay-figure` ist
`width: 100%`, wodurch die Box überbestimmt ist und der negative
rechte Margin verworfen wird.
* Foto-Paare stellten beide Bilder gleich breit dar, wodurch ein
Hochformat neben einem Querformat doppelt so hoch wurde. Jetzt eine
Justified-Reihe: Breite proportional zum Seitenverhältnis, dadurch
exakt gleiche Höhen ohne Beschnitt.
* Die Lightbox hatte im Journal keine Tastatursteuerung und keine
Scroll-Sperre — beides lag in PhotoGrid, also nur im Album-Raster.
Verlagert in die Lightbox selbst, der doppelte Handler in PhotoGrid
entfernt (sonst wäre pro Pfeildruck zwei Bilder weitergesprungen).
Zusätzlich sind Proofing-Herzen und das EXIF-Panel im redaktionellen
Modus abschaltbar und im Journal aus: Client-Proofing ist ein
Übergabe-Workflow für Alben, keine Funktion einer Reportage. Auf der
Übersicht wurde der Untertitel doppelt gerendert, weil extractExcerpt()
auf ihn zurückfällt.
* feat(admin): eigene URLs für alle Bereiche und überarbeiteter Journal Studio
Das Panel war eine einzige Client-Komponente, die ihre Tabs über
useState umschaltet: keine Route pro Bereich, kein Deep-Link, kein
Zurück-Button, und nach jedem Reload landete man wieder auf "Pages".
Auch der Journal-Editor hatte keine eigene URL.
Jetzt sieben echte Routen (/admin, /admin/pages, /admin/journal,
/admin/journal/<slug>, /admin/settings, /admin/settings/<section>,
/admin/analytics). Die Shell mit Auth-Prüfung und Kopfzeile sitzt im
Layout und umschließt children — so bleibt die Session-Prüfung über
Routenwechsel montiert, statt bei jedem Tab-Wechsel neu zu laufen und
den Spinner aufblitzen zu lassen. Unbekannte Settings-Sections fallen
auf General zurück.
Im selben Zug der Journal Studio, der die Routing-Änderung ohnehin
berührt:
* Alle Eingabefelder nutzten die Klasse `admin-input`, die in keinem
CSS definiert war — sie fielen auf Browser-Defaults zurück. Jetzt
gestylt, gescoped, damit der generische Name nicht ins übrige Panel
leckt.
* Der Split View ist per Divider verschiebbar (Maus, Tastatur,
Doppelklick zum Zurücksetzen), Breite in localStorage.
* Das Studio bricht aus der 1120px-Spalte aus, die Preview hat Platz.
* Die Preview zeigte für jedes Foto fest 3:2 und log damit über das
Layout der veröffentlichten Seite; die echten Verhältnisse werden
jetzt an den Thumbnails gemessen.
* Blöcke mit alten positionsbasierten Foto-Referenzen werden markiert —
die lassen sich auf einer eigenständigen Journal-Seite nicht auflösen
und verschwanden bisher stillschweigend.
* Blocktyp-Chips als gemeinsame Komponente, monochrom und per Icon
unterschieden statt in einer eigenen Blau/Lila/Orange-Palette, die
sonst nirgends im Panel vorkommt.
* Emojis raus, SVG-Icons rein.
* fix(admin): Designsystem-Lücken im Panel schließen
Eine systematische Suche nach Klassen, die im TSX benutzt, aber in
keinem CSS definiert sind, hat vier Stellen mit demselben Muster
gefunden:
* `admin-btn-secondary` und `admin-btn-danger` existierten nicht.
Sämtliche Löschen-Buttons sahen aus wie normale Buttons — keine
Warnwirkung. Als `.admin-btn.admin-btn-danger` geschrieben, weil die
Varianten mit den Größenklassen kombiniert werden, die weiter unten
ihren eigenen Hintergrund setzen.
* `page-progress-fill` fehlte: die Balken im Analytics-Panel waren
leere Rinnen, der Wert unsichtbar. Ebenso `analytics-empty` und
`admin-loading-container`.
* `subpage-drawer-section`, `subpage-drawer-title-group` und
`input-with-icon` im PageBuilder waren wirkungslos und teils mit
Inline-Styles notdürftig kompensiert.
* `grid-card-preview` fehlte, wodurch alle Layout-Vorschauen auf Höhe
null kollabierten. Nur "Showcase" war sichtbar — und das aus Versehen,
weil sein Tile `hero-tile` hieß und damit die `aspect-ratio` einer
fremden PageBuilder-Klasse erbte. Umbenannt.
Dazu die Theme-Vorschauen, die praktisch identisch aussahen: alle
sieben zeigten dieselben drei quadratischen Kacheln, einziger
Unterschied war die Hintergrundfarbe. Sie spiegeln jetzt die echten
Preset-Daten aus lib/config/theme.ts — Schriftcharakter, Eckradius,
Rasterdichte, Passepartout bzw. Schatten.
Die Kopfzeile richtet sich an einer gemeinsamen Steuerhöhe aus. Der
"Site"-Link war höher als die Buttons daneben: `.admin-btn` hatte keine
explizite line-height, und <button> bekommt box-sizing: border-box vom
Browser, <a> nicht. Die Statusanzeige setzte weder Schriftfamilie noch
-größe und fiel auf den Browser-Default zurück — sie war nie Teil des
Designsystems. Jetzt gleiche Höhe und Typografie, Pillenform als
Zustandsanzeige, Trenner zur Aktionsgruppe. Das dauerhafte Pulsieren
des grünen Punkts entfällt; Animation bleibt den Zuständen vorbehalten,
die Aufmerksamkeit brauchen.
Restliche Emojis durch SVG-Icons ersetzt.
* feat(admin): add scroll-to-top toggle and SVG arrow
The back-to-top button was always on and drew its arrow with the ↑ text
character. Add a `scrollToTop` setting (General section in the admin
panel, defaults to on so existing sites are unaffected) and replace the
glyph with a stroked SVG that inherits currentColor across all presets.
Scroll handling is now rAF-throttled, and prefers-reduced-motion turns
off both the hover lift and the smooth scroll.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* security(journal): escape author HTML instead of filtering it
CodeQL flags 26 high-severity alerts on the release PR; the bulk of them
point at sanitizeHtml(), a denylist of three regexes whose output goes
into dangerouslySetInnerHTML. I reproduced the bypasses before changing
anything — all four pass the old filter:
<img src=x onerror=alert(1)> unquoted handler, regex wants quotes
<img src=x onerror='alert(1)'> single quotes, same reason
<svg onload=alert(1)> not a <script> tag
javasjavascript:cript:alert(1) one-pass replace, the rest recombines
Plus an attribute breakout in the link renderer, which interpolated the
URL into href="${url}" unescaped:
[x](" onmouseover=alert(1) x=") -> <a href="" onmouseover=alert(1" ...>
Replaced with escaping. Every `&<>"'` in author text becomes an entity
before any markdown rule runs, so the only tags in the output are the
ones renderInlineMarkdown builds itself and a URL cannot leave its
attribute. Link schemes are now an allowlist (http, https, mailto, tel,
relative, anchor); anything else renders as plain text.
The serializer needed decodeHtmlEntities to match: without it a save
would write `<` into the Markdown file and the next parse would
escape the `&` again, corrupting the text further on every round trip.
A test covers parse -> serialize -> parse stability.
Two existing tests asserted the old behaviour and were rewritten rather
than deleted: one expected the script tag to be *removed* from the
output, the other expected the substring "onerror" to be absent. Both
now assert the property that matters — that no tag can form — because
`onerror` surviving as inert text is fine.
Exploitation required admin access either way: journal and essay text
comes from the studio or from content/*.md. This removes the weakness,
it does not close an open door.
* fix(journal): contain file paths, fix frontmatter escaping, defuse two regexes
Follow-up to the CodeQL findings on release PR #432. I went through all 23
alerts individually; this addresses the ones that are worth addressing and
documents the ones that are not.
Path containment (12 alerts, all false positives). Every path built from a
slug goes through isValidSlug(), a strict `/^[a-zA-Z0-9_-]+$/` allowlist —
verified against `../../etc/passwd`, `foo/../../bar`, `a/b`, `..` and
friends, all rejected. CodeQL cannot see it because the check lives in
another module. Added a containedPath() helper that resolves the path and
refuses anything landing outside the content directory: a second lock on a
door that was already locked, and a barrier the analysis can follow. Also
removed the dead `slug.endsWith('.md')` branches — isValidSlug rejects
dots, so they could never be taken.
Frontmatter escaping (1 alert, a real bug). The serializer escaped `"` but
not `\`, and the parser never unescaped anything. A title containing either
character gained a backslash on every save/load cycle:
Path C:\Users "quoted" -> Path C:\\Users \"quoted\" -> ...
Escaping now covers the backslash first and the parser undoes both. A round
trip test covers quotes, backslashes, colons and apostrophes.
Two of three ReDoS regexes (2 alerts). The heading pattern had `\s+` next to
`(.+)`, which overlap, so the engine tried every split of the whitespace run
before failing; it is `[ \t]+` and `(\S.*)` now. The quote author pattern
wrapped a lazy group around an optional group — the quadratic case — and is
an index split on the first separator now.
The third regex, the bold/italic pass, is deliberately left alone. Rewriting
it with a negated class removes the backtracking but breaks nesting:
`**bold *and italic* here**` renders as `*<em>bold </em>and italic<em> here</em>*`.
The input is Markdown from an authenticated admin, never a visitor, so the
worst case is an author slowing down their own page. The comment in the code
says so.
Not addressed: 5 js/insufficient-password-hash alerts. They ask for a slow
KDF where the code uses SHA-256/HMAC — correct in general, but the admin
password comes from an environment variable and is never stored, so a
stronger KDF protects nothing that is not already in plaintext in the
environment.
399 unit tests pass (18 new), build and tsc clean.
* fix(journal): split the quote author by scanning, not by pattern
Cleaning up after myself. #439 replaced the quote author regex
`/^(.*?)(?:\s+--\s+(.+))?$/` — a lazy group around an optional one, the
quadratic case — with `/\s+--\s+/`, and CodeQL promptly flagged the
replacement for the same reason. One reported pattern traded for another.
splitQuoteAuthor() scans for the first `--` that has whitespace on both
sides. Linear, and it needs no reasoning about the regex engine at all.
Behaviour is identical to the original regex, checked case by case
against it: separator, no separator, several separators, tabs, wide gaps,
`a--b` without spaces, leading and trailing separators, empty string. All
nine are covered by tests, plus one asserting it stays fast on a 50k
whitespace run.
* security(admin): derive the session signing key with scrypt
getSigningKey() ran the admin password through a single SHA-256 digest.
It is key derivation rather than password storage, so it was never a
stored-hash problem — but if the derived key ever leaked, one fast hash
makes recovering the password by brute force cheap. CodeQL flags it as
js/insufficient-password-hash and, for this particular alert, has a
point.
scrypt now. Two details that matter:
The salt has to be stable. A random salt would produce a different key on
every call and invalidate every session; it is derived from AUTH_SECRET,
which is deployment-specific. Verified across separate processes: same
inputs produce a verifying token, so a restart or redeploy does not log
everyone out.
The result is cached for the process lifetime. scrypt is deliberately
slow and getSigningKey() runs on every admin request — deriving per call
would add ~24ms to each one. The cache key is a digest of secret plus
password, so a rotation misses the cache and derives again, and no extra
plaintext copy of the password is kept around.
Measured: first derivation 24ms, subsequent calls 0.12ms. Password
change and secret change both invalidate an existing token, checked in
separate processes.
Note the one-time effect: existing admin sessions are invalidated by the
change in derivation, so everyone logs in once more.
* security(admin): key the signing-key cache on its inputs, not a digest
#441 removed the SHA-256 digest of the admin password from the key
derivation and introduced a new one three lines above it, as the cache
key. Same rule, same file: CodeQL reported it immediately.
The comment justifying it was wrong too. It claimed no second plaintext
copy of the password was retained, while building the digest input
allocates exactly that in order to hash it.
The cache now holds the two input strings and compares them directly.
They are the same objects the env and install caches already hold, so
nothing extra is retained, and no fast hash of the password exists in
memory for the life of the process.
Re-verified after the change, in separate processes: a token from one
process verifies in another with the same inputs, a changed password or
secret rejects it, first derivation 24ms, cached calls 0.001ms.
That leaves scrypt as the only place the password is hashed here. The
two HMACs further down operate on the derived key and on token data, not
on the password; the constant-time comparison in verifyAdminPassword is
the previously dismissed one.
---------
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: ralksta <ralksta@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: lancetm714 <108419425+lancetm714@users.noreply.github.com>
2026-08-14 16:55:07 +02:00
|
|
|
|
INSTALL_CONTENT_DIR: process.env.INSTALL_CONTENT_DIR || undefined,
|
2026-03-08 21:12:52 +01:00
|
|
|
|
};
|
2026-02-14 11:47:26 +01:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/** Validated, typed environment variables. */
|
|
|
|
|
|
export const env = parseEnv();
|